a simple introduction to ads/cft and its application to ... · a simple introduction to ads/cft and...
TRANSCRIPT
A simple introduction to AdS/CFT
and its application to condensed matter physics.
D-ITP Advanced Topics in Theoretical Physics Fall 2013
K. Schalm and R. Davison1
1Instituut-Lorentz for Theoretical Physics,
Universiteit Leiden, P. O. Box 9506,
2300 RA Leiden, The Netherlands∗
(Dated: December 8, 2013)
Abstract
The anti-de Sitter/ Conformal Field theory correspondence provides a unique novel perspective on critical
phenomena at second order quantum phase transitions in systems with spatial dimensions d > 1. The first
half of these lectures will provide technical background to apply the so called ”holographic” techniques of
the correspondence. The second half discusses the application to quantum phase transitions in condensed
matter: how spontaneous symmetry breaking in a quantum critical system is similar and different to the
standard case, the notion of semi-local quantum liquids and their connection to non-Fermi liquids and
strange metals.
There are also many lecture notes available. A sample of references are:
• J. Erdmenger, Introduction to gauge gravity duality, Chapters 1,2,4,5,6.
• S.A. Hartnoll, Lectures on holographic methods for condensed matter physics, Class. Quant. Grav.
26, 224002 (2009).
• N. Iqbal, H. Liu and M. Mezei, Lectures on holographic non-Fermi liquids and quantum phase tran-
sitions.
∗Electronic address: [email protected]
1
Contents
I Lecture I
1. Context and Background 4
2. Anti-de-Sitter space 5
3. Exercises 8
4. The AdS/CFT correspondence 10
4.1. CFT brush up 10
4.2. AdS/CFT 10
II Lecture II:
5. The physics of AdS/CFT Correlation functions 11
6. AdS/CFT at finite temperature: Black Holes, and the Hawking Page
transition as the holographic encoding of confinement/deconfinement 14
7. AdS/CFT and hydrodynamics: minimal viscosities 17
7.1. Fluctuation dissipation theorem and Kubo relations 19
8. Exercises 24
III Lecture III:
9. Condensed Matter in a nutshell 25
9.1. The Landau Fermi Liquid Theory as an IR stable fixed point. 29
9.1.1. Spectral Functions 31
9.2. DC conductivity in Landau Fermi liquid theory 31
9.2.1. Macroscopic features of FL follow from Quasi 1D 33
9.2.2. Luttinger’s theorem 34
2
9.3. Experimental non-Fermi-liquids 35
10. Finite density holography/Semi-local quantum liquid/AdS2-metal 38
10.1. AdS2 near horizon geometry and emergent local quantum criticality 40
11. Exercise: Computing the conductivity of the AdS-RN metal 41
11.1. The promise of AdS/CFT 46
IV Lecture IV: Holographic superconductors, electron stars
and semi-local quantum liquids
12. Holographic non-Fermi liquids 47
12.1. Fermionic correlation functions from holography 47
12.1.1. On various Green’s functions... 52
12.2. Non-Fermi liquids from holography 52
13. Holographic superconductors 54
13.1. (In)stability analysis 56
13.2. The holographic superconductor solution 57
13.2.1. The macroscopic properties of the holographic superconductor 59
14. Exercises 61
References 61
3
Lecture I
1. CONTEXT AND BACKGROUND
In QFT (merging of QM & SR) particles are localized excitations of fields. Localization
=Assumes weak coupling. In strong coupling regime, wave functions start to overlap until
indistinguishable. Point: there should be other (non-perturbative) excitations that correctly
describe the physics in this regime. Can we identify them? If so we have an example of a
duality. The difficult part is the identification. There are no hard and fast rules to do so. In
some cases we know and we will illustrate this with a simple example of electric-magnetic
duality in d = 3+1 dimensions. It is extremely simple as opposed to the topic of this course,
AdS/CFT, which will be technically very involved. At its heart philosophically AdS/CFT
is just another duality, however.
Example: Electromagnetic duality
S[A] =
∫d4x− 1
4g2F 2µν (1.1)
Bianchi identity
∂[µFνρ] = 0 ⇒ Fνρ = ∂[νAρ] (1.2)
EOM
∂µFµρ = Jρelectric (1.3)
Introduce magnetic charge
∂[µFνρ] = εµνρσJσmagnetic (1.4)
Symmetric. Really dual, means one can also think of Bianchi as e.o.m. Can be made
manifest for J = 0 by using a Lagrange multiplier
S[F, A] =
∫d4x− 1
4g2F 2µν − Aµεµνρσ∂νF ρσ (1.5)
Note now Fµν is the field. Integrating out (i.e. solving the e.o.m. for) the Lagrange multiplier
Aµ gives us back the original action.
∂S
∂Aµ= εµνρσ∂
νF ρσ = 0 ⇒ Fνρ = ∂[νAρ] (1.6)
4
However, we can also integrate out Fµν . It is now algebraic. Completing squares
S[A] =
∫− 1
4g2F 2µν + ∂νAµεµνρσ∂
νF ρσ (1.7)
=
∫− 1
4g2(Fµν − g2εµνρσ∂
ρAσ)2 +g2
4εµνρσ(pa[ρAσ])2 =
∫−g
2
4(∂[ρAσ])
2 (1.8)
where we used that εµνρσεαβγδηργησδ = −2(ηµαηνβ − ηµβηνα). If we define g = 1/g this is
the original action. Note however that perturbation expansion in g is the strong-coupling
expansion in g and vice versa. This is the simplest example of a (strong-weak) duality.
AdS/CFT is just an incredible more complicated version of the same story, where the dual
theory is now given by a quantum-gravitational string theory in a curved space with one extra
spatial dimension. Nevertheless a precise dictionary exists between these higher-dimensional
gravitational theories and the original CFT. That dictionary and some of its uses is the goal
of these lectures.
2. ANTI-DE-SITTER SPACE
The AdS in AdS/CFT stands for anti-de-Sitter space. Here we give a brief introduction
to this spacetime. As always in physics, the best way to classify spacetimes is through
their symmetries. For spacetimes symmetries are the isometries of the spacetime manifold
i.e. coordinate transformations that leave the metric invariant. To each such symmetry
we can therefore associate a distinct Killing vector. Recall that under general coordinate
transformations δxν = ξν(x) the metric transforms as
δgµν = D(µξν) (2.1)
A Killing vector is a vector such that D(µξν) = 0.
The simplest spacetimes are those with the most symmetries. These will have the most
Killing vectors. Since a d-dimensional symmetric metric has d(d + 1)/2 components, there
can be at most d(d+1)/2 independent Killing equations and thus at most d(d+1)/2 Killing
vectors. Spacetimes with such amount of Killing vectors are called maximally symmetric.
These Killing vectors will form a group. The simplest group that contains this number of
generators of SO(d + 1), but this would not account for the Lorentzian signature. It turns
out there are only three distinct maximally symmetric spacetimes. They are classified by
the group formed by their Killing vectors. They are
5
Group Spacetime
SO(d, 1) d-dimensional de Sitter space
SO(d− 2, 2) d-dimensional anti-de Sitter space
ISO(d− 1, 1) = SO(d− 1, 1) ∝ Translationsd d-dimensional Minkowski space
The simplicity of maximally symmetric spacetimes reflects through in their curvature. Since
essentially every point is similar to every other point the curvature cannot have a derivative
dependence. Thus1
Rµνρσ = c(gµρgνσ − gµσgνρ) (2.2)
Rµν = c(d− 1)gµν (2.3)
R = cd(d− 1) (2.4)
with c a constant. All these spacetimes are in fact solutions to the vacuum dynamical
Einstein equations supplemented with a cosmological constant
Rµν −1
2gµν(R− 2Λ) = 8πGTµν (2.5)
In vacuum Tµν = 0, then a contraction with the metric gives
R =2d
d− 2Λ (2.6)
and thus c = 2Λ/(d− 1)(d− 2).
It turns out that Λ > 0 corresponds to de Sitter space, Λ = 0 to Minkowski space,
and Λ < 0 to anti-de-Sitter space. In other words, anti-de-Sitter space is the maximally
symmetric space that is the unique solution to the vacuum Einstein equations with a negative
cosmological constant.
This may still not say much, but you are in fact very familiar with maximally symmetric
Euclidean spaces. The maximally symmetric Euclidean spaces are the sphere (Λ > 0),
flat space (Λ = 0), and the hyperboloid (Λ < 0). de Sitter space is thus the Lorentzian
generalization of the sphere and anti-de Sitter space is the Lorentzian generalization of the
hyperboloid. Just as the d-dimensional sphere is defined as the solution to the constraint
X21 +X2
2 + . . .+X2d+1 = R2 (2.7)
1 We are using conventions where [Dµ, Dν ]V ρ = R ρµν σV
σ and gµν is mostly plus.
6
. and the hyperboloid is defined as
−X2−1 +X2
1 + . . .+X2d = −R2 (2.8)
d-dimensional anti-de Sitter space is defined as the
−X2−1 −X2
0 +X21 +X2
d−1 = −R2 (2.9)
We immediately see the SO(d− 1, 2) symmetry. A solution to this defining relation is
X1 = R+ cos θ1 (2.10)
X2 = R+ sin θ1 cos θ2 (2.11)
... = (2.12)
Xd−2 = R+ sin θ1 sin θ2 . . . cos θd−2 (2.13)
Xd−1 = R+ sin θ1 sin θ2 . . . sin θd−2 (2.14)
X0 = R− cos t (2.15)
X−1 = R− sin t (2.16)
where R2+ = R2 sinh2 τ, R2
− = R2 cosh2 τ . The induced metric one gets is
ds2 = −dX2−1 − dX2
0 + dX21 + . . . dX2
d−1 (2.17)
= R2(dτ 2 − cosh2 τdt2 + sinh2 τdΩ2d−2) (2.18)
where dΩ2d−2 is the metric on the d− 2-dimensional sphere Sd−2. Anti-de-Sitter space is the
universal cover of this, where we unroll the periodic coordinate t. This coordinate system
is quite special because it is a global coordinate system. It covers the whole of AdS (recall
that generically one needs multiple coordinate patches to cover the whole space).
Now consider the coordinate transformation τ = arcsinh tan ρ with 0 ≤ ρ ≤ π/2. This
maps the hyperbolic direction τ to a finite range. In these coordinates the metric is
ds2 =R2
cos2 ρ
(dρ2 − dt2 + sin2 ρdΩ2
d−2
)(2.19)
This metric allows us to understand the topology of AdS. For this we can ignore the overall
conformal factor R2/ cos2 ρ. The topologically equivalent spacetime
ds2teq ∼
(dρ2 − dt2 + sin2 ρdΩ2
d−2
)(2.20)
7
To be filled
FIG. 1: Topology of anti-de Sitter space.
describes a cylinder with radial direction ρ, longitudinal direction t and each point in (ρ, t)
is an Sd−2. See Fig.
We thus see that AdS has a (conformal) boundary at ρ = π/2. Note that in physical
units this is an infinite distance away. This conformal boundary will play an important role
in AdS/CFT.
There is a third convenient coordinate system for AdS. See exercises that it indeed solves
the defining relation (2.9)
ds2 =R2
z2
(dz2 − dt2 + dx2
1 + . . .+ dx2d−2
)(2.21)
This coordinate system, the so-called Poincare patch, covers only half of AdS. Relating it
back to the global coordinate system through the defining relation (2.9) one can show that
it covers the diamond depicted in Fig 1.
For a review see [106, 107].
3. EXERCISES
Problem 1:
In lecture we showed electromagnetic duality in 3 + 1 dimensions. In 2 + 1 dimensions the
exact same exercise is called particle-vortex or Abelian-Higgs duality. Starting from a free
U(1) gauge theory in 2 + 1 dimensions perform the same steps to dualize the theory. What
is the theory you end up with, i.e. what type of theory is the vortex/Higgs theory?
Problem 2:
a. In lecture we claimed that the spacetime defined the embedding
−X2−1 −X2
0 +X21 + . . .+X2
d−1 = −R2 (3.1)
8
into RRd−1,2 can be chosen to have local coordinates
X1 = R+ cos θ1 (3.2)
X2 = R+ sin θ1 cos θ2 (3.3)
... = (3.4)
Xd−2 = R+ sin θ1 sin θ2 . . . cos θd−2 (3.5)
Xd−1 = R+ sin θ1 sin θ2 . . . sin θd−2 (3.6)
X0 = R− cos t (3.7)
X−1 = R− sin t (3.8)
where R2+ = R2 sinh2 τ, R2
− = R2 cosh2 τ gives the global AdS metric. Verify this.
b. There is another well-known choice to solve the AdS constraint as as an embedding.
It is given by
Xi = Rxi/z i = 0, . . . , d− 2 (3.9)
Xd−1 +X−1 =1
z(x2
0 − x21 − . . .− x2
d−2 − z2) (3.10)
Xd−1 −X−1 = R2 1
z(3.11)
What does the AdS metric look like for this choice?
Problem 3: In lecture we computed the two-point function of a scalar conformal operator
from AdS using the AdS/CFT prescription. Do the same for a vector operator. There are a
number of subleties. (1) As mentioned, in AdS the vector field is dynamical. The only way
such a field makes sense is as a gauge field. We must therefore fix the gauge the solve for
the Green’s function. (2) Because of gauge invariance the vector field is massless. This will
make the solution a bit simpler.
a. Write down the equation of motion for a vector field in an AdSd+1 background.
b. Verify that the bulk-boundary propagator
Gµν =zd−2
(z2 + xµxµ)d−1ηµν (3.12)
G0µ = − zd−3xµ(z2 + xµxµ)d−1
(3.13)
9
with all other components vanishing solves the equation of motion. Show also that it
reduces to a delta-function on the boundary.
c. Consider an arbitrary boundary source aµ(xν) and its corresponding bulk solution
AM =∫GMνa
ν . Substitute this into the bulk action and use the AdS/CFT dictionary
to determine the two-point function of conformal vector fields.
d. Verify that these conformal vector fields are conserved (global) currents.
A brief concluding comment. The gauge choice we have used here is not a standard one:
it is one where the Green’s function in position space is particularly simple. In more in-
volved AdS/CFT computations, especially numerical, one normally chooses the where the
components along the extra AdS direction vanish, i.e. Az = 0.
4. THE ADS/CFT CORRESPONDENCE
4.1. CFT brush up
A CFTd is a special relativistic d-dimensional QFT which is invariant under SO(d, 2)
instead of just the Lorentz group SO(d− 1, 1).2 Now think of renormalization of a standard
QFT. This is a very deep statement that in the real world microscopic physics resonates
through for macroscopic physics, but in a very predictable way that does not depend on the
details of the microscopic model. A generic QFT is therefore scale dependent. CFTs by
definition, however, are scale independent. This must mean that CFTs do not renormalize:
all beta-functions and anomalous dimensions must be exactly zero.
4.2. AdS/CFT
〈e∫ddxJ(x)O(x)〉CFT =
∫Dφ e−SAdS
∣∣∣∣φ(x,∂AdS)=J(x)
. (4.1)
2 Non-relativistic CFTs also exist, but we will not consider them here.
10
boundary: field theory bulk: gravity
energy momentum tensor T ab metric field gab
global internal symmetry current Ja Maxwell field Aa
order parameter/scalar operator Ob scalar field φ
fermionic operator Of Dirac field ψ
spin/charge of the operator spin/charge of the field
conformal dimension of the operator mass of the field
source of the operator boundary value of the field (leading part)
VEV of the operator boundary value of radial momentum of the field
(subleading part)
(Global aspects)
global spacetime symmetry local isometry
temperature Hawking temperature
chemical potential/charge density boundary values of the gauge potential
phase transition Instability of black holes
TABLE I: The basic dictionary for AdS/CFT correspondance.
Lecture II
5. THE PHYSICS OF ADS/CFT CORRELATION FUNCTIONS
Let us now show how the GKPW rule precisely validates our ad hoc construction Eq.(??)
that the two-point correlation function can be read off from the near-boundary asymptotics
of AdS waves. The GKPW rule instructs us to consider the on-shell action with the boundary
value of the field equal to the source in the dual field theory. For the simple scalar theory
the action can be written in terms of a “bulk” and “boundary” contribution as,
S = −1
2
∫AdS
dD+1x√−gφ(+m2)φ− 1
2
∮∂AdS
dDξ√−hφ∂nφ (5.1)
Here h is the determinant of the induced metric hab = gµν∂aξµ∂bξ
ν on the boundary which
has the local coordinates ξa(x). Only the bulk part contributes to the equations of motion.
On-shell, i.e. when we substitute for φ(x) a solution for to the equation of motion, the
11
“bulk” term integrated over the whole of the AdS space vanishes. We already learned that
this solution — i.e. with appropriate boundary conditions in the interior — has a universal
asymptotic behavior near the boundary z = 0 as,
φsol(ω, k, z) = A(ω, k)z∆− +B(ω, k)z∆+ + . . . . (5.2)
According to GKPW the boundary value of φ(ω, k, z) is proportional to the source J . A
priori we are now facing a problem since ∆− will be generically negative and thus the
boundary value of φ is not well defined. However, we know what the meaning of this
divergence is in the boundary field theory. Approaching the boundary is like increasing
the renormalization scale to infinity and here one typically encounters UV divergences. In
other words, the theory has to be regulated and this can be done in a particularly elegant
way using the bulk language. GKPW proposed that one should compute at an infinitesimal
distance z = ε away from the formal boundary, and then modify the theory such that one
can take an appropriate limit ε→ 0. Let’s do so. The “regulated” on-shell action equals
Son-shell(ε) =1
2
∮z=ε
dωdD−1k
(2π)Dz−D+1
(∆−A
2z2∆−−1 + (∆− + ∆+)ABz∆−+∆+−1 + . . .). (5.3)
The first term is formally divergent. The key to making the action well-defined is that adding
an arbitrary boundary term to the action never changes the equation of motion. However,
such an extra boundary term can be used to remove by hand the UV divergence. Adding a
boundary counterterm of the form
Scounter(ε) = −1
2∆−
∮z=ε
dωdD−1k
(2π)D
√−hφ2
= −1
2∆−
∮z=ε
dωdD−1k
(2π)Dz−D
(A2z2∆− + 2ABz∆++∆− + . . .
)(5.4)
yields in combination with Eq. (5.3),
Son-shell(ε) + Scounter(ε) =1
2
∮z=ε
dωdD−1k
(2π)Dz−D
((D − 2∆−)ABzD + . . .
). (5.5)
This is now all finite. We can now equate the leading behavior (A coefficient) of the φsol with
the source J . Given that the above should coincide with the combination iJ〈O〉 in the field
theory, by taking the single derivative with respect to J this yields the expectation value
〈O〉 of the field theory operator sourced by J in AdS/CFT in the presence of the source. It
is given by
〈O(ω, k)〉J = 2νB(ω, k) (5.6)
12
where we used ∆± = D2± ν. Thus we see that equating the leading near-boundary behavior
A(ω, k) of the solution φsol with the source implies that the subleading near-boundary behavior
B(ω, k) is the corresponding response.
One can similarly obtain the two-point correlation function by taking an additional deriva-
tive w.r.t. J and setting it to vanish. Linear response theory tells us already that B(ω, k)
ought to be proportional to A(ω, k) and the proportionality is precisely the CFT Green’s
function,
〈O(−ω,−k)O(ω, k)〉 = 2νB(ω, k)
A(ω, k)(5.7)
and we have demonstrated that the propagator rule Eq.(??) is indeed a consequence of the
fundamental GKPW rule.
One can actually check that the GKPW rule encodes linear response theory by itself
in a correct fashion. For this one has to realize that one in essence just needs to solve a
simple Dirichlet boundary value problem. Recall that the equation of motion for φ has two
independent solutions. Let us denote the solution with A = 0 as φB with boundary behavior
φB(z) = Bz∆+(1 +∑
n cnzn). This is the appropriate Dirichlet solution that vanishes at the
boundary, and let us denote as φint(z) the solution with boundary behavior determined by
regularity in the interior of AdS as in Eq. (??). The Dirichlet AdS Green’s function obeying
limz→0 GAdS = 0 thus equals
GAdS(z, z′) =φB(z)φint(z
′)θ(z − z′) + φint(z)φB(z′)θ(z′ − z)
φint∂φB − φB∂φint. (5.8)
The Wronskian in the denominator assures the correct normalization and is independent of
z, i.e. it may be evaluated for any preferred z. Then for a boundary source J(ω, k) the
solution to the equation of motion is
φsol(ω1, k1, z) = limε→0
∮z′=ε
dωdD−1k
(2π)D∂z′G(z, ω1, k1; z′, ω, k)J(ω, k)
= limε→0
∫dωdD−1k
(2π)D∂φB(ε)φint(z)
φint∂φB − φB∂φintJ(ω, k). (5.9)
By construction this obeys limz→0 φsol(ω, k, z) = J(ω, k), which can be seen by noting that
the Wronskian reduces to φint∂φB for z → 0. The normal derivative of the solution follows
13
straightforwardly
∂zφsol(ω1, k1, z) = limε→0
∮z′=ε
dωdD−1k
(2π)D∂z∂
′zG(z, z′)J(ω, k)
= limε→0
∫dωdD−1k
(2π)D∂φB(ε)∂φint(z)
φint(ε)∂φB(ε)J(ω, k) = lim
ε→0
∫dωdD−1k
(2π)D∂φint(z)
φint(ε)J(ω, k)
(5.10)
Substituting this into the action one finds
Son-shell + Scounter = limz→0
(1
2
∫dDxz−D+1φsol∂zφsol −
1
2∆−
∫dDxz−Dφ2
sol
)= lim
ε→0
∫dωdD−1k
(2π)D
(ε−D+1 1
2J(−ω,−k)
∂φint(ε)
φint(ε)J(ω, k)− ∆−
2ε−DJ(−ω,−k)J(ω, k)
).
(5.11)
Near the boundary the solution φint has again the generic behavior φint = Aint(ω, k)z∆− +
Bint(ω, k)z∆+ and by taking two derivatives w.r.t. the source J one finds
〈O(−ω,−k)O(ω, k)〉 = limε→0
ε−D+1∂φint(ε)
φint(ε)− 1
2∆−
= ε−D+1∂ε(ε−∆−φint(ε))
φint(ε)= 2ν
B(ω, k)
A(ω, k). (5.12)
Thus one is left with the same linear response answer (5.7).
6. ADS/CFT AT FINITE TEMPERATURE: BLACK HOLES, AND THE HAWK-
ING PAGE TRANSITION AS THE HOLOGRAPHIC ENCODING OF CONFINE-
MENT/DECONFINEMENT
The Einstein equation for a D + 1 dimensional Minkowski-signature space time with a
negative cosmological constant is,
Rµν −1
2gµνR−
D(D − 1)
2L2gµν = 0. (6.1)
This is solved by an AdS-Schwarzschild black hole background with the metric,
ds2 =r2
L2(−f(r)dt2 + dx2
i ) +L2
r2f(r)dr2, i = 1, ..., D − 1 (6.2)
where L is the AdS radius, r is the radial direction, and the redshift factor
f(r) = 1− rD0 /rD. (6.3)
14
gives us the black horizon at r0. To compute the temperature we use an insight from Gibbons
and Hawking. Consider a general static black hole metric
ds2 = −gtt(r)dt2 +dr2
grr(r)+ gxx(r)d~x
2,
where gtt(r) and grr(r) have a single zero at the horizon r0, and Wick rotate to the Euclidean
signature τ = it with the result
ds2E = gtt(r)dτ
2 +dr2
grr(r)+ gxx(r)(d~x)2. (6.4)
Let’s make the natural assumption that the properties of the black hole are reflected in the
geometry near the horizon where gtt and grr are vanishing. To focus in on this region we
expand gtt(r) = g′tt(r0)(r − r0) + ..., grr(r) = grr ′(r0)(r − r0) + ... and gxx(r) = gxx(r0) + ...
where the prefactors g′’s are just numbers. Thus the near horizon (Euclidean) metric equals
ds2E = g′tt(r0)(r − r0)dτ 2 +
dr2
grr ′(r0)(r − r0)+ gxx(r0)(d~x)2 + . . . (6.5)
It is now convenient to re-parametrize the radial direction in terms of a new variable R =
2√r − r0/
√grr ′ and the metric becomes
ds2E =
1
4R2g′ttg
rr ′dτ 2 + dR2 + gxx(0)(d~x)2 + . . .
What matters is the plane spanned by this R and the imaginary time direction τ . This is just
like the metric of a plane in polar coordinates with τ being the compact angular direction.
Upon approaching the horizon R → 0 one sees that the prefactor of dτ 2 is vanishing: this
means that the Euclidean time direction shrinks to a point! However, since the horizon is
not a special point, we should not allow this point to be singular. Smoothness at the horizon
can be achieved by insisting that R = 0 is the center of a Euclidean polar coordinate system
and this implies that τ is periodic with period 4π√g′tt(r0)grr ′(r0)
. This periodicity is directly
identified with the inverse temperature of the black hole (as measured at r =∞).
For the AdS-Schwarzschild black hole we thus find
T =
√g′ttg
rr ′
4π=
r20
4πL2
df(r)
dr|r=r0 =
Dr0
4πL2. (6.6)
Let us illustrate the above with the most important equations for the bulk. As we
emphasized, to introduce a second scale in addition to the temperature we make the spatial
15
topology compact. All Euclidean isotropic solutions to AdS-Einstein’s equations have the
same generic form as Eqn. (??)
ds22 = f(r)dτ 2 +
dr2
f(r)+ r2dΩ2
D−1, (6.7)
The most general solution is
f(r) = 1 +r2
L2− ωD
M
rD−2, ωD =
2κ2
(D − 1)Vol(SD−1). (6.8)
For M = 0 we have “thermal AdS” (the time circle is still compact); for M 6= 0 we have the
Euclidean (global-)AdS black hole.3
Note that f(r) now has multiple zeroes. The outer most one, i.e. the larger solution of
the equation
1 +r2
L2− ωD
M
rD−2= 0 (6.10)
is the horizon r+. Following the near-horizon prescription of the previous section, one can
deduce the temperature
β =4πL2r+
Dr2+ + (D − 2)L2
. (6.11)
in units of c = kB = ~ = 1 and its inverse gives the temperature of the black hole T = 1/β.
We can compare which solution is thermodynamically favored by using the algorithm
of the previous section to compute the free energies of the boundary corresponding with
these two different bulk geometries. A crucial extra ingredient is that in order to compare
these free energies, one has to insist that the bulk geometries describe precisely the same
boundary space-time. Choosing a cut-off in the radial direction at r = R, one can express
the imaginary time periodicity of the thermal AdS β′ in terms of the black hole inverse
temperature β such that both systems live in the same boundary space, sharing the same
boundary time circle.At a cut-off radius R the periodicities in each geometry are
β′(1 +
R2
L2
)1/2= β
(1 +
R2
L2− ωDM
RD−2
)1/2. (6.12)
3 In the scaling limit
t = λt, r = λ−1r, dΩ2D−1 = λ2d~x2 with λ→ 0, (6.9)
the metric (6.7) changes to the vacuum planar AdSD+1. If we scale M → λ−DM at the same time, one
gets the planar AdS Schawartzshild black hole (6.2).
16
One now computes the difference in the boundary field theory free energy straightforwardly
and it is related to the Euclidean action difference
βFThAdS − β′(R)FBH =d
κ2L2limR→∞
(∫ β
0
dt
∫ R
r+
dr
∫SD−1
dΩrD−1 −∫ β′
0
dt
∫ R
0
dr
∫SD−1
dΩrD−1
)=
4πVol(SD−1)rD−1+ (L2 − r2
+)
2κ2(Dr2+ + (D − 2)L2)
. (6.13)
where FThAdS and FBH are the free energies for the black hole- and thermal AdS solutions.
With the units restored, we have
β(FThAdS − FBH) = c3 4πVol(SD−1)rD−1+ (L2 − r2
+)
2κ2(Dr2+ + (D − 2)L2)
. (6.14)
7. ADS/CFT AND HYDRODYNAMICS: MINIMAL VISCOSITIES
One step further then equilibrium thermodynamics is to consider small time-dependent
fluctuations around a thermodynamic equilibrium. For very small energies and ultra-long
wavelengths the state is still collective and the response of the system is described by hydro-
dynamics. A slightly more detailed definition would say that hydrodynamics applies when
the length scale of interest is much larger than the mean-free-path between the microscopic
constituents of the system. This is where the connection with AdS/CFT comes in. Since
the mean-free-path is inversely proportional to the coupling constant, for strongly coupled
theories hydrodynamics applies widely. And AdS/CFT can give a gravitational description
of (matrix-valued) strongly coupled theories (in the large N limit).
The defining equations of relativistic hydrodynamics are extremely simple. They are just
conservation of energy-momentum and charge
∂µTµν = 0∂µJ
µ = 0 (7.1)
These are d + 1 equations for 12d(d + 1) + d unknowns. And the dynamical fluid equations
must always be supplemented by additional constraints. The fluids you are probably familiar
with are perfect fluids. These fluids are completely homogeneous and isotropic (rotationally
and translationally invariant) and experience no force, i.e. they move at a constant velocity.
Call this velocity uµ, normalized such that uµuµ = −1. Then the rest frame expression for
17
the stress-tensor in terms of the energy density ε and pressure p
T µν =
ε 0 . . . 0
0 p . . . 0
0 0 . . . 0
0 0 . . . p
(7.2)
together with the symmetries uniquely determine the stress-tensor of a perfect fluid to be
Tµν = (ε+ p)uµuν + pgµν (7.3)
The extra relation one must add is the equation of state which determines p in terms of ε.
In reality no perfect fluid exists. One should think of it as one thinks of an ideal gas. In
reality interactions at the microscopic scale have an effect on the expression for the stress-
tensor that causes it to deviates from the perfect fluid. Note that the symmetries played an
essential role in the definition of the perfect fluid. So what is true is that the more and more
these symmetries are manifested in the system, the better the perfect fluid description is.
Vice versa this means that deviation from the perfect fluid are captured by terms that
break the perfect translational and rotational symmetries of the system. The fluid velocity
thus has small variations in space and time. For simplicity let us consider the fluid in the
rest frame, perform a lowest order expansion in gradient velocities and arrange these again
according to representations of the rotational group. One has
T00 = ε , T0i = 0 (7.4)
Tij = εδij + η
(∂iuj + ∂jui −
2
3∂`u
`
)+ ζδij∂`u
` (7.5)
A relativistic expression exists, and can be found in many references (e.g. Baier). Here
η and ζ are the shear and bulk viscosities. They are decay constants how rotations and
decompressions relax.
For simplicity and physics let us consider the charge current rather than the energy-
momentum current. The insight is that a small local excess of charge density should cause
a small current to flow away from the excess until it is completely diluted into the system.
Improving on the perfect fluid limit one would write down in the rest frame
J0 = ρ , J i = −D∂iρ = −D∂iJ0 (7.6)
18
with D the diffusion constant. This second constitutive equation is also known as Fick’s
law. Now take the divergence of both sides
∂iJi = −DJ0 (7.7)
and use the dynamical equation (the charge conservation law) ∂µJµ = 0 to write this as
∂0J0 = DJ0 (7.8)
Solving this in Fourier space (space only) one has
J0(t, k) = e−Dk2tJ0
init(k) (7.9)
One sees the diffusion of charge happen, as the excess decays away over time. The shear and
and bulk viscosities play the same role as the diffusion constant D with some more indices
to carry around.
7.1. Fluctuation dissipation theorem and Kubo relations
Let’s now also Fourier transform the charge diffusion equation for J0 in time.
iωJ0(ω, k) = Dk2J0(ω, k) ⇔ (iω −Dk2)J0(ω, k) = 0 (7.10)
One should think of this dynamical equation as the equation of motion for the charge density.
Suppose now one would like to solve a more complicated problem in this fluid, e.g. one with
an external source for the charge. Then one would construct the Green’s function from the
equation of motion
G = 〈J0J0〉 ' 1
iω −Dk2+ . . . (7.11)
(Recall there are higher order terms in the gradient expansion. Since we are doing semiclas-
sical physics, this has to be the retarded Green’s function). This means that
ImiG =Dk2
ω2 +D2k4(7.12)
and one can get D from by taking considering
limω→0
ImiG =1
Dk2(7.13)
19
This Kubo relation is an example of the fluctuation-dissipation theorem. One can get macro-
scopic properties (the diffusion constant) by considering microscopic fluctuations. A mani-
festion of the fluctuation-dissipation theorem that you almost certainly know is the optical
theorem: the total cross section is proportional to the imaginary part of the forward scat-
tering amplitude.
Let us know use AdS/CFT to compute the diffusive properties of a strongly coupled
(matrix)-theory. The example that we will use is the shear viscosity [48]. This is for two
reasons (1) historically it was the first, (2) the finding that the viscosity is universal and
bounded from below for all theories with a gravitational dual, (3) but, most importantly,
in units of the entropy density it is the smallest viscosity known and it turned out to be in
very close neighborhood to viscosities measured of the Quark-Gluon-Plasma in Relativistic
Heavy Ion Collisions and Cold Fermi-gas experiments. We can do it here because it is among
the simplest calculations for a linear response property one can imagine in the AdS/CFT
context. The dynamics in the bulk is just pure Einstein gravity, while the geometry is
that dual to thermal equilibrium: an AdS Schwarzschild black holeOne of the aspects this
calculation will illustrate is that in AdS/CFT it is remarkably straightforward to compute
directly in real time in thermal systems.
We will compute the viscosity from the Kubo formula
η = − limω→0
1
ωImGR
x1x2,x1x2(ω, 0), (7.14)
where, GRx1x2,x1x2
(ω, 0) is the retarded Green’s function of the xy component of the energy
momentum tensor, defined as
GRµν,αβ(ω,~k) = −i
∫dtd~xeiωt−i
~k·~xθ(t)〈[Tµν(t, ~x), Tαβ(0, 0)]〉. (7.15)
Using the GKPW formula the stress tensor perturbations are encoded in fluctuations of the
metric. As usual in linearized gravity, one writes the metric as gµν = g0µν + hµν where g0
µν
is the background metric that solved for the AdS Schwarzschild black hole while hµν is the
infinitesimal metric fluctuation, i.e. the gravitational wave/graviton.
The counting of the number of DOF for the massless graviton is a bit of a tricky affair:
We focus on 4 + 1 dimensional bulk. We first classify the graviton modes as follows. We
take the spatial momentum to be in the x3 direction ~k = (0, 0, k) and this means that the
perturbation hµν = hµν(t, r, x3). The system has an SO(2) symmetry in the x1x2-plane
20
and according to the behavior of the graviton modes under this symmetry we have three
decoupled sets of graviton modes: the tensor mode (transverse mode) hx1x2 , the vector
mode (shear mode), a linear combination of htx1 , hx3x1 , hrx1 together with htx2 , hx3x2 , hrx2
and two scalar modes (sound mode) from two linearly independent combinations of htt, htx3 ,
hx3x3 , hx1x1 + hx2x2 , hrr, htr and hrx3 . It directly follows from the linearized coordinate
transformations on hµν
δhµν = ∂µεν + ∂νεµ (7.16)
that hx1x2 in this momentum configuration is a gauge-invariant, i.e. physical mode.
We now perturb the Einstein equation with a negative cosmological constant, eqn (6.1),
around the solution of the AdS-Schwarzchild black hole geometry (6.2). Denoting the per-
turbation hx1x2(t, r, x3) ≡ gx1µhµx2 with [47]
δgx1x2 = hx1x2(t, r, x3) =
∫dωdk
(2π)2φ(r;ω, k)e−iωt+ikx3 , (7.17)
one obtains linearized equation of motion for φ(r;ω, k). One finds that it identical to the
equation for a massless scalar
1√−g
∂µ√−ggµν∂νφ(r; t, x3) = 0 (7.18)
Substituting the background metric eq. (6.2) and Fourier transforming we find
φ(r;ω, k)′′ +
(D + 1
r+f ′
f
)φ′(r;ω, k) +
(ω2 − k2f)L4
r4f 2φ = 0 (7.19)
with f = (1− rD0rD
).
To obtain the retarded Green’s function in the CFT we must now solve this equation with
infalling boundary conditions at the horizon. The retarded Green’s function is then given
by the ratio of the subleading to the leading coefficient of the solution at the boundary:
φsol(u, ω, k) = A(ω, k)r−∆− +B(ω, k)r∆−−D + . . . (7.20)
with ∆− = 0 in this case, and
GCFTR,xy,xy(ω, k) ∼ B(ω, k)
A(ω, k)(7.21)
To obtain A(ω, k) and B(ω, k) is a relatively straightforward numerical exercise; in some
special cases the answer is analytically known. Rather than using this brute force short-cut,
21
we will give an alternate way [53, 54] to compute the solution which captures some of the
essential physics. We first write the Green’s function directly in terms of the solution φsol
following the GKPW construction
GCFTR (ω, k) =
1
2κ2limr→∞
r−2∆−√−ggrr ∂rφsol(r)
φsol(r)
(7.22)
The overall factor 1/2κ2 follows from the normalization of the Einstein-Hilbert action.
The key step is that the viscosity (and all other transport coefficients) follow from the
imaginary part of the retarded Green’s function. Inserting 1 = φ∗(r)/φ∗(r) inside the limit
one obtains for this imaginary part the expression
ImGCFTR (ω, k) = lim
r→∞r−2∆−
√−ggrrφ
∗sol(r)∂rφsol(r)− φsol(r)∂rφ
∗sol(r)
2iφ∗sol(r)φsol(r)(7.23)
The numerator is readily recognized as the Wronskian which measures the flux density
through a surface at fixed r. The physics is that for a field obeying a second order equation
of the type
φ′′ + P (r)φ′ +Q(r)φ = 0 (7.24)
with P (r) and Q(r) real, the generalized Wronskian
W (r) = e∫ r P (r)(φ∗sol(r)∂rφsol(r)− φsol(r)∂rφ
∗sol(r)) (7.25)
is conserved: ∂rW = 0. Note that the combination√−ggrr is precisely this necessary
prefactor: this is readily seen by acting with ∂r and using the equation of motion (7.18).
Rewriting the imaginary part of the retarded Green’s function directly in terms of the
conserved Wronskian we find
ImGCFTR (ω, k) = lim
r→∞r−2∆−
W (r)
2iφ∗sol(r)φsol(r)(7.26)
We will discuss the physics of this rewriting in a moment. Mathematically its compu-
tational power for frequency-independent transport coefficients is immediate. We can use
the conservation of the Wronskian to evaluate the numerator at any point r. The most
convenient is the horizon itself where the infalling boundary conditions are set. Near the
horizon, near the single zero of f(r), the equation of motion (7.19) reduces to
φ′′ +f ′
fφ′ +
L4ω2
r4f 2φ+ . . . = 0 (7.27)
22
Writing f(r) = (r − r0)Dr0
+ . . .,
φ′′ +1
(r − r0)φ′ +
L4ω2
D2r20(r − r0)2
φ+ . . . = 0. (7.28)
We can deduce the powerlaw dependence of the solution near the horizon by substituting
the ansatz
φsol(r;ω, k) = (r − r0)α(1 + . . .) (7.29)
One finds
α(α− 1) + α +L4ω2
D2r20
= 0 (7.30)
with solutions α = ± iωL2
Dr0= ± iω
4πT. In the last step we used the relation between the
horizon location and the black hole temperature derived in eqn. (??). The choice α =
−iω/4πT corresponds to the infalling solution. Thus near the horizon we may parametrize
φsol(z;ω, k) = (r − r0)−iω/4πTF (r;ω, k), (7.31)
where F (r;ω, k) is regular at the horizon r = r0. Evaluating now the conserved Wronskian
near the horizon one finds
W (r0) = limr→r0
√−ggrrφ∗
↔∂ φ
= limr→r0
rD+1
LD+1
(1− rD0
rD
)(−2iω
4πT(r − r0)−1
)F ∗(1)F (1) + . . .
=rD+1
0
LD+1
D
r0
−2iω
4πTF ∗(1)F (1)
= (4
DπTL)D−1(−2iω)F ∗(1)F (1). (7.32)
In the last line, we have again used the definition of the temperature 4πTL = Dr0/L.
Substituting, one sees that the remaining unknown in the Green’s function
ImGCFTR (ω) =
1
2κ2limr→∞
(4
DπTL)D−1(−ω)
F ∗(1)F (1)
F ∗sol(r)Fsol(r)(7.33)
is the ratio of the absolute value of F (r) at the horizon to |F (r)| at the boundary. Formally
one still needs to solve for F (r) to determine this. However, in the limit ω → 0, k → 0 the
leading contribution will be the ω-independent solution for the remaining function F . From
23
equation (7.19) this is readily seen to be the trivial constant function. This is the leading
solution φ ∼ Ar−∆− near r →∞ with our special case of ∆− = 0. Hence one obtains
ImGCFTR (ω, k) =
1
2κ2(
4
DπTL)D−1(−ω). (7.34)
From the Kubo relation
η = − limω→0
1
ωIm〈Tx1x2(−ω)Tx1x2(ω)〉 = − lim
ω→0
1
ωImGCFT
R (ω, 0), (7.35)
the shear viscosity therefore equals
η =1
2κ2(
4
DπTL)D−1. (7.36)
Recalling that 2κ2 ≡ 16πG and comparing this to the entropy density (??) s =
4π2κ2
( 4DπTL)D−1 for a D + 1-dimensional AdS Schwarzschild black hole, we find the famous
ratio
η
s=
1
4π
~kB. (7.37)
8. EXERCISES
24
Lecture III
9. CONDENSED MATTER IN A NUTSHELL
Condensed matter is about systems at finite density. When you put together a large
amount of matter, how does it behave? For relatively dilute systems at finite temperature,
one can readily apply the lessons of statistical physics. Many interesting phenomena occur,
however, at low temperatures when quantum effects become important and/or when one
considers densities large enough that the quantum wavefunctions of each of the constituents
can overlap. This is the regime we will investigate here: we wish to know the macroscopic
characteristics — density, pressure, equation of state, etc. — of T/µ → 0 low temperature
(low energy) finite density quantum matter.
There is a large set of condensed matter wisdoms which we have acquired over decades
that tell us what generically can happen when you study such quantum matter.
• The system can be gapped. This means that all excitations cost a finite amount of
energy. In relativistic language, there is no massless state. Therefore at T = 0, i.e.
when there is no energy density in the system, nothing (interesting) in fact happens.
The system is in its groundstate, but is unable to respond to infinitesimal small (in
energy) perturbations.
• The groundstate of the system can spontaneously break a global symmetry. This is the
generic groundstate for a system of bosons. In the spontaneously broken or ordered
state, the system has protected gapless/massless Goldstone modes. These modes then
completely dominate the low temperature/low energy physics. The formalism to de-
scribe this is the Landau-Ginzburg free energy functional: a low-energy effective action
for the order parameter of the symmetry breaking. From the free energy we can get
all the other macroscopic equilibrium properties of the system.4 The robustness of the
4 Note that compared to particle physics, these Landau-Ginzburg functionals are often very simple. One
only concentrates on the order parameter and not on any other degrees of freedom in the system. Thanks to
Goldstone’s theorem this is almost always sufficient at extremely low energies and macroscopic equilibrium
properties. Particle physics Lagrangians are usually much more involved. One also wishes to know detail
the behavior of excitations at moderate energies. In condensed matter one mostly wishes to connect
the microscopics to the correct Landau-Ginzburg functional. In these broken groundstates, one therefore
makes a moduli-space approximation.
25
results are guaranteed by Goldstone’s theorem. In high dimensions the usual scaling
arguments reduce the Landau-Ginzburg functional to a Gaussian (mean-field approx-
imation). In lower dimensions quantum effects (fluctuations) are more important and
one can get distinct non-mean field behavior.
• For a system with only fermions, one generically gets a Landau Fermi liquid. This is
based on the Pauli principle. Recall that we are interested in matter at finite density
and take a free fermion as the simplest example of a many-body-fermion system. In this
free Fermi gas, two fermions cannot be in the same quantum state. They therefore
must have different momentum. The occupied states of the fermions fill shells in
momentum space outward from the origin as each shell has a slightly larger energy
cost. The last fermion added to the system this way gives a preferred momentum
scale, the fermi-momentum kF (Fig 1.). The Fermi liquid is this Fermi gas state where
one includes small interactions between the Fermions.
The Fermi liquid has a well-defined stable fermionic quasiparticle excitation that costs
no energy to excite. It is massless and it therefore controls the low-temperature/low-
energy macroscopic behavior of the system. Its free energy is essentially that of a free
non-interacting fermion. The robustness of the Fermi liquid follows from the Pauli
principle. Intuitively this is very understandable. Strictly speaking, however, this
is mathematically not as well understood as the Goldstone theorem for bosons. A
way to phrase this is that there is no order parameter language for the Fermi liquid
state. The closest mathematical explanation for the robustness is Luttinger’s theorem.
For charged fermions this states that indepedent of the nature of the interactions the
total charge carried by the Fermi liquid state equals the volume of occupied states
in momentum space expressed in terms of kF We will discuss Luttinger’s theorem in
detail below.
This is what happens generically. The primary quest of quantum matter is to understand
those systems that do not fall into one of these three classes.
A more refined version of this question is as follows. A compressible state is a finite
density state of a system with a U(1) symmetry, in which the charge Q as a function of the
26
chemical potential µ behaves as
〈Q〉 ≡ − ∂
∂µF ∼ µα , α > 0 (9.1)
In particular if a locally change the chemically potential, I will locally change the charge
density. This state therefore conducts. In condensed matter nomenclature, it is therefore a
metal. The amount of energy does not feature in this argument, so at infinitesimally small
energies, this will still happen, and therefore a compressible state does not have a gap.
Recapitulating the lessons from above with one addition, the known compressible states
are:
• A state with broken translational symmetry, i.e. a crystal. The state is a solid.
• A state with spontaneously broken U(1) symmetry. One has a superconductor with
a single massless Goldstone mode. Note that technically we have broken a global
U(1) and not the local U(1) of Maxwell theory. In many condensed matter situations
the electromagnetic interaction is so weak, however, that the mean free path of a
photon is much larger than the sample size. This allows one to effectively think of the
electromagnetic current as a global U(1) current. At the end one can weakly gauge
the global U(1) to incorporate dynamical electromagnetic effects. From this viewpoint
the spontaneously broken state is a superconductor, rather than a superfluid.
• The state is a Fermi-liquid. This is special in that it is not associated with a broken
symmetry.
The question then is:
Is there a translationally invariant compressible state of matter with-
out symmetry breaking that is not a Fermi liquid, and how do we
characterize it?
This is where AdS/CFT will shed new light.
First, however, we will study in particular the Fermi liquid in some more detail. This
will also explain why this question is so important. Before we do so, note that a finite
density systems is manifestly associated with a U(1) charge. This connection points to
a strongly identifying macroscopic characteristic of the state, in addition to the pressure,
27
density, etc.The expectation value of the current Jµ associated to this U(1) current is an
important macroscopic characteristic of this state. The timelike component J0 is just the
charge density. In equilibrium the expectation value of the spacelike component vanishes,
of course, as it is odd under time reversal. But its fluctuations also contains very important
information. To compute these, couple to an external source Aexternali as usual and add this
to the action
〈J i〉 = 〈 δ
δAexternali
S〉 (9.2)
We are already argued that in the absence of a driving source the expectation value
vanishes. Thus for a small but finite source one finds
〈J i〉 = 〈 δ
δAexternali
S〉 = 〈J iJ j〉Aexternalj +O(A) (9.3)
One can think of this in extremely physical terms. We know that an external electric field
can source a current. Specifically, in a gauge where A0 = 0, ∂tAi = Ei. In Fourier space this
becomes Ai(ω) = Ei(ω)/(−iω). Now the amount of the current that this external electric
field generates is controlled by the conductivity σ(ω) of the system. A larger conductivity
results in a larger current etc. For a not-too-large electric field this is expressed through
the defining relation 〈J i(ω)〉 ≡ σij(ω)Ei(ω) where σij(ω) is the conductivity tensor. The
frequency dependent conductivity is also known as the optical or AC conductivity. The
ω → 0 limit of the optical conductivity is the DC conductivity. Combining this defining
relation with the current as a response to an external source, one obtains the expression for
the conductivity
σij(ω) =∂
∂Ej(ω)〈J i(ω)〉 =
1
−iω∂
∂Aexternalj (ω)
〈J iJk〉Aexternalk
=−1
iω〈J i(ω)J j(ω)〉 (9.4)
Thus for small fields one can read off the conductivity by measuring the two-point cor-
relation of the current. Because the conductivity is closely related to the U(1) charge of
the chemical potential introduced to put the system at a finite density, we will encounter it
extensively throughout these lectures.
28
9.1. The Landau Fermi Liquid Theory as an IR stable fixed point.
Before we start using AdS/CFT, for reference we present a modern view of Landau
Fermi liquid theory. As we already alluded briefly, the Landau Fermi liquid is based on the
zero temperature free Fermi gas. The momentum of the highest occupied state is called
the Fermi momentum kF and this momentum will characterize the macroscopic properties
of the system. For charged free fermions the number density is equal to the U(1) charge
density. The finite density system is therefore obtained by introducing a chemical potential
for the U(1)-charge. The low-energy (=low-frequency) dynamics are then readily seen to be
described by expanding the free (non-relativistic spinless) electron action coupled to a U(1)
gauge field
S =
∫ddxΨ†(x)(i(∂t − ieAt) +
1
2m∇2)Ψ(x) (9.5)
in the presence of a finite chemical potential background
At = µ/e
S =
∫ddxΨ†(x)(i∂t + µ+
1
2m∇2)Ψ(x) (9.6)
One readily notes that there is a preferred momentum — the Fermi momentum kF =√
2mµ — for which the kinetic term action vanishes at ω = 0, i.e. the state at k = kF costs
zero energy to excite. The low energy effective action is the expansion around this state
[112, 113].
SFermi gas =
∫ddk
(2π)dΨ†(k)(i∂t − vF (k − kF ))Ψ(k) + . . . (9.7)
with vF = kF/m. Conventionally one invokes the Pauli principle that turning on small
interactions cannot in any way change this occupation barrier and the existence of a mo-
mentum scale kF , and postulates that the relevant degrees of freedom are still the fermions
above. In the modern view this insight directly follows from the renormalization group
[112, 113]. Assuming that the interactions between the fermions are controlled by a scale
M , one can integrate out the interactions to arrive at an effective theory for the low-energy
fermions. In perturbation theory this results in analytic higher order corrections to the dis-
persion relation plus induced interactions. As long as the scale M is much much larger than
29
the chemical potential, rotational invariance uniquely determines
SFermi liquid =
∫ddk
(2π)dΨ†(k)
(i∂t + µ+
∇2
2m+
α
M(i∂t + µ)2 +
β
M(i∂t + µ)
∇2
2m+
γ
M
(∇2
2m
)2
+ . . .
)Ψ(k)
+ξ
Md(Ψ†Ψ)(Ψ†Ψ) + . . .
=
∫ddk
(2π)dΨ†(k)
(ω − vF (k − kF ) +
α
Mω2 +
β
Mω(k − kF ) +
γ
M(k − kF )2 . . .
)Ψ(k)
+ξ
Md(Ψ†Ψ)(Ψ†Ψ) + . . . (9.8)
where α, β, γ, ξ are dimensionless coupling constants generically of order O(1).
Straightforward dimensional analysis shows that all interactions between the electrons
are irrelevant in any dimension d > 1. The generic fixed point of an interacting finite
density fermi system is thus a field theory of IR free fermionic quasiparticles. This generic
effective low-energy theory is the Fermi liquid. Its tell-tale sign is that the two-point retarded
correlation function of two such quasiparticles
GR(ω, k) =1
ω − vF (k − kF )− Σ(ω, k)(9.9)
=1
ω − vF (k − kF )− Σ′(ω, k)− iΣ′′(ω, k)(9.10)
has a self-energy Σ(ω, k) whose imaginary part ImΣ(ω, k) ≡ Σ′′(ω, k) behaves as
Σ′′(ω, kF ) ∼ αMω2 + . . . for k = kF . This is the reflection of the analyticity of the gra-
dient expansion. The spectral density
A = − 1
πImGR(ω, k) (9.11)
= − 1
π
Σ′′(ω, k)
(ω − vF (k − kF ) + Σ′(ω, k))2 + (Σ′′(ω, k))2(9.12)
therefore has a beautiful Lorentzian line shape around ω = 0 for k = kF .
A =1
π
αω2/M
(ω − vF (k − kF ))2 + α2 ω4
M2
+ . . . (9.13)
with total weight unity and width α/M . At finite temperature T the usual t = −iτ ,
ω = iωE with the Matsubara frequencies ωE = nT , means that “dimensionally” we can
obtain a lot of information by just replacing ω → ω + iT in the imaginary part of the
self-energy.
30
9.1.1. Spectral Functions
Single Fermion spectral functions play an important role in condensed matter physics.
This is readily understandable as Fermi liquid theory is the foundation of band theory in
crystals and as shown above the spectral function of a Fermi liquid has a distinct Lorentzian
peak at ω = 0, k = kF . This peak can be directly measured with Angle Resolved Photo
Emission Spectroscopy (ARPES). The probability of liberating an electron from a sample by
shining light (photoemission) is directly proportional to the spectral function. Particularly
in the last decade the resolution of energy and momentum of this final state electron has
improved so much that extremely detailed spectral functions can be measured. The con-
ventional way to express the single fermion spectral function is to first expand the Green’s
function (9.9) around ω = 0 and k = kF as
GR(ω, k) =1
ω − vF (k − kF )− Σ′(0)− ω∂ωΣ′(0)− iΣ′′(ω) + . . .(9.14)
=Z
ω − mm?vF (k − kF )− iΓ
+Gincoherent (9.15)
Here Z ≡ 1/(1 − ∂ωΣ′(0)) is known as the pole strength Γ = ZΣ′′(ω) the width, and
m? = mvFkF/(vFkF − Σ′(0)) the renormalized mass. [Work out further]
9.2. DC conductivity in Landau Fermi liquid theory
As an example of how the low-energy degrees of freedom determine the macroscopic
properties, we will now compute the temperature dependence of the DC conductivity of the
Fermi liquid. We will do so using the “Kubo” relation derived earlier
σ(ω) =−1
iω〈 ~J ~J〉 (9.16)
In the effective action for the Fermi liquid, the current J i is the composite operator
J i = ie2m
Ψ∂iΨ (there are no gamma-matrices, as we are considering a spinless fermion for
simplicity). The two point function of the currents to leading order is therefore the one-loop
31
diagram
(9.17)
We will now simply state the result for the real part of this diagram. The real part can
be powerfully rewritten in terms of Fermi-Dirac distributions f(ω) and the spectral function
A(ω, k). In dS spatial dimensions one has
Re(σDC) ∼ limω→0
1
iωImΨΨΨΨ
∼ limω→0
1
iω
∫ddSk
dω1
2π
dω2
2π
f(ω1)− f(ω2)
ω1 − ω − ω2 − iεA(ω1, k)Λ(ω1, ω2, ω,−k)A(ω2, k)Λ(ω2, ω1, ω, k)
(9.18)
Essentially this formula states that a photon with energy ω can create an electron pair
with energies ω1 and ω2 pulled from the occupied states f(ωi)A(ωi, k). The relative minus
sign between f(ω1) and f(ω2) follows from Fermi statistics and all the model depedent
details are stuck in the transport vertex Λ(ω1, ω2, ω, k). We next make the assumption that
this vertex is in fact momentum independent (at low energies). This is not entirely obvious,
but can be verified to be so in most microscopic models. To leading order in ω Λ is then a
constant, and we get
σDC ∼ limω→0
1
iω
∫ddSk
dω1
2π
dω2
2π
f(ω1)− f(ω2)
ω1 − ω − ω2 − iεA(ω1, k)A(ω2, k)
∼∫ddSkdω1
df
dω1
A(ω1, k)2 (9.19)
One can get the dominant temperature-dependence in the spectral function by shifting
the the imaginary part of the self-energy Σ′′(ω)→ Σ′′(ω + iT ). This lifts the pole at ω = 0,
but nevertheless the spectral function A(ω, k) remains dominated by the Fermi-surface at
ω ∼ 0, k ∼ kF .
One now notes that the Fermi surface is a dS − 1-dimensional surface in momentum
space with a single normal direction k⊥. Because of the Pauli principle to lowest order all
excitations can only carry momentum in the direction k⊥. This means that dS − 1 of the
momentum integrals computing the DC conductivity localize on the Fermi surface. Doing
32
so
σDC ∼ kdS−1F
∫dk⊥dω
df
dωA(ω, k⊥)2 (9.20)
we see that the system has reduced to an effective d = 1 system. Next we note that
the derivative of the Fermi-Dirac distribution is essentially a delta-function at ω = 0, so we
can perform the ω-integral as well. We can now scale the temperature dependence out of
integral by a redefinition k⊥ → T 2k⊥/M and we deduce characteristic temperature-squared
dependence of the conductivity of a Fermi-liquid.5 [CHECK]
σDC ∼ kdS−1F
∫dk⊥A(0, k⊥)2|Σ′′(0+iT ) (9.21)
∼ kdS−1F
∫dk⊥
T 4/M2((−vFk⊥)2 + T 4
M2
)2 (9.22)
∼ T 2
T 4kdS−1F M ∼ 1
T 2kdS−1F M (9.23)
Experimentally one usually measures the resistivity ρ. This is just the inverse of the
conductivity. Thus for a Fermi liquid
ρFL ≡1
σFLDC∼ T 2 (9.24)
9.2.1. Macroscopic features of FL follow from Quasi 1D
Completely following our general introduction, all the macroscopic properties of the Fermi
liquid are essentially explained by this effective dimensional reduction of the massless quasi-
particle to a dS = 1 dimensional dynamics perpendicular to the Fermi surface.
For instance for a general massless system in dS dimensions where there is no intrinsic scale
the temperature scaling of the entropy density and the specific heat is simply determined
by dimensional analysis.
s ∼ 1
LdS∼ T dS
vdS(9.25)
cV ∼T
∂S∂T ∼ T dS
vdS(9.26)
This is called hyperscaling. Here v is the characteristic IR velocity that appears in the
dispersion relation of the massless excitation ω ∼ vk. For a relativistic system clearly v = c.
5 Note that the DC-conductivity is dimensionless in dS = 2 as it should.
33
For a Fermi liquid the dS-dimensional system reduces to deffS = 1 dimensional system
due to the presence of the Fermi surface at kF . As there is no other scale in the system, one
can assume that hyperscaling holds in this reduced sense and argue that the temperature
scaling of the entropy and specific heat behave as
s ∼ T
k
ds−1
F∼ kdSF
T
EF(9.27)
cV ∼T
EF(9.28)
This very simple insight is indeed borne out by a detailed microscopic calculation.
9.2.2. Luttinger’s theorem
Let us finally briefly discuss Luttinger’s theorem. In the Fermi liquid it is very easy
to establish a relation between the number density and the Fermi momentum kF . The
expectation value of the number density as a function of momentum is
nF (k) =
∫dω
2πi〈Ψ†(ω, k)Ψ(ω, k)〉 (9.29)
Again the right hand side is a two point correlation function and to lowest order this
therefore can be computed from the one-loop expression
nF (k) = limt→0−
∫dω
2πiGR (9.30)
= limt→0−
∫dω
2πi
Z
ω − vFk⊥ − iεsign(ω)(9.31)
= Zθ(kF − k) (9.32)
This is the classic step function profile of the zero-temperature occupation density of a
free Fermi gas. The total number density for the free Fermi gas therefore equals
ntotal =
∫d2k
(2π)2Zθ(kF − k) = Z
k2F
4π(9.33)
(In the free system Σ = 0, but we need to be explicit about the iε prescription in that
case.) For this free system the number density is directly proportional to the total charge
density in the system. In the conventional normalization
Q ≡ 4π2qntotal = πk2F (9.34)
34
This relation is Luttinger’s theorem. Its power relies in the fact that Luttinger has
proved that no perturbative or non-perturbative interactions can change this relation.[Is
this because conserved currents do not renormalize?]. There is a caveat. The Fermi
liquid groundstate must remain, e.g. turning on a BCS instability which causes pairing and
condensation of the composite pair operator can violate this theorem.
As mentioned, Luttinger theorem is in some way the Fermi liquid equivalent of Gold-
stone’s theorem. It explains the robustness of the Fermi surface and its associated massless
quasiparticle. Nevertheless there is no fundamental understanding of Luttinger’s theorem
as spontaneous symmetry breaking provides for Goldstone’s theorem.
9.3. Experimental non-Fermi-liquids
So far the story may seem quite academic. The importance of the question whether there
are compressible states of matter without symmetry breaking that are not a Fermi liquid
became immediate with the discovery of high temperature superconductors in 1985. These
fall outside the standard BCS paradigm. Within BCS theory, one can derive a gap equation,
which gives the value of the order parameter as a function of the temperature
∆(T ) = ΛUV e− 1λnF (T ) (9.35)
[CHECK] Here ΛUV is a UV cut-off, λ is the small perturbative coupling parameter that
controls the binding energy of the Cooper pair and nF is the number density. The identifying
characteristic is the exponential suppression. This precludes BCS superconductivity at any
high temperature. For normal materials one has, and one finds an upper bound for the
critical temp
Tmaxcritical BCS = K (9.36)
High temperature superconductors violate this bound by definition. The mechanism
underlying superconductivity can therefore not be the conventional BCS mechanism. The
superconducting phase itself is actually not the mystery. As explained superconductivity is
just the spontaneous breaking of U(1) symmetry and this can has no temperature limitations.
Indeed experimentally one readily shows that one has (d-wave) Cooper paired electrons that
Bose condense. The mystery is why the material becomes superconducting at such a high
35
temperature. The failure of BCS theory means that it cannot be a regular Fermi liquid
with a perturbative Cooper pairing interaction. Indeed experimentally the macroscopic
characteristics of the strange metal state just above the critical temperature are not those
of a regular Fermi liquid. The most notable of these is the fact that the resistivity scales
linear in the temperature rather than quadratic.
The typical phase diagram of a high temperature superconductor is given in Fig. 2. At
optimal doping — the doping value where Tc is maximal — the material is in this strange
metal phase for T > Tc and superconducts for T < Tc. For higher doping above Tc there is
a crossover from the strange metal to a regular Fermi liquid regime. Very near the doping
where the superconducting region vanishes, the onset of superconductivity can in fact be
explained by standard BCS theory. For lower than optimal doping at temperatures higher
than Tc there is a phase transition to another mysterious pseudogap phase6 and at even lower
doping a crossover to an antiferromagnetic phase.
FIG. 2: The typical phase diagram of a high Tc superconductor.
Our focus here will be on the strange metal phase. Quite soon after the discovery of
high Tc superconductivity, a very phenomenological proposal was made to explain the linear
resistivity [135]. Suppose the full single fermion spectral function would be of the form
G =1
ω − vFk⊥ + iω lnω(9.37)
6 Only very recently has it been settled that this is a true phase transition, see [? ? ].
36
From our derivation of the scaling behavior of the resistivity, it is easy to say that if
Σ′′ ∼ ωa, then the conductivity scales as σ ∼ T−a and thus the resistivity as ρ ∼ T a.
The logarithm is needed to differentiate it analytically from the true linear term in ω.
This “model”, or rather this Green’s function, is clearly not that of a regular Fermi liquid.
Collectively (phenomenological) single fermion spectral functions that do not have a width
that scales as ω2 are called non-Fermi liquids. And this particular one with Σ′′ ∼ ω lnω is
called the marginal Fermi liquid.
Although a tremendous effort has been made to understand the strange metal better and
beyond the phenomenology of the marginal Fermi liquid, only in one qualitative aspect has
progress been made. It is now believed that the barrier to understanding is explained by
the fact that underlying the physics of the strange metal is the phenomenon of Quantum
Criticality. This is the critical universal behavior that occurs in the vicinity of a Quantum
Phase transition. The latter is a phase transition which happens at exactly T = 0; it is
driven by quantum rather than thermal fluctuations. Now the relevant aspect is that if
this phase transition is second order, the absence of a scale at the critical point (due to
the diverging correlation length) means that the quantum field theory describing this point
must be a conformal field theory. The special aspect of a quantum critical theory compared
to a classical critical theory is that one now raises the temperature in the conformal field
theory, the conformal constraints resonate through in the finite temperature physics. A
CFT at finite T is still very special in that all its dynamics are still controlled by the T = 0
conformal symmetry and in general the only aspect that changes is that all dimensionfull
quantities are now given in terms of the only present scale T . To get non-conformal (generic)
behavior one needs at least two scales. This means that the phase diagram near a quantum
critical point looks as in Fig. 9.3 The “fan-like” region is indeed very suggestive when
compared to the phase diagram of high Tc superconductors. Although the idea that the
physics underlying the strange metal is a finite T conformal field theory, in detail it is not
so simple. In particular scale-invariance is only observed in terms of energy-temperature
scaling. In spatial directions one still notes a distinct Fermi surface with ARPES data. This
curious combination, scale-less in the “time-direction”, but a distinct Fermi-momentum in
the spatial directions has been coined “local quantum criticality”.
The Fermi-surface and Fermions apparently continue to play a key role in the strange
metal region. An approach to understanding the strange metal and its quantum criticality
37
g
T
gc
0
InsulatorSuperfluid
Quantumcritical
TKT
Monday, December 21, 2009
FIG. 3: Generic phase diagram of a quantum phase transition
is to understand the role of fermionic degrees of freedom in strongly interacting CFTs. That
will be our ultimate goal. First we shall try to understanding strongly interacting CFTs,
i.e. quantum critical systems, at finite density more generally.
10. FINITE DENSITY HOLOGRAPHY/SEMI-LOCAL QUANTUM
LIQUID/ADS2-METAL
To describe a system at finite (charge) density we recall that global charges in the CFT,
correspond to local symmetries on the AdS gravity side. Thus the minimal system we need
to consider is AdS-Einstein gravity — dual to the energy-momentum sector — plus Maxwell
theory — dual to a U(1) global charge. Moreover, from statistical physics we recall that we
can think of the chemical potential as the “source” for charge density. Since the expectation
value of the current is given by the subleading part of the solution to Maxwell’s equation,
we deduce that an AdS system dual to a field theory at finite chemical potential will be a
solution to the AdS equations of motion with the other leading non-normalizable solution
to Maxwell equation non-vanishing.
The simplest such solution is the charged AdS black hole. In the remainder of this
discussion of AdS/CFT we will consider AdS theories in d = 3 + 1 dimensions, describing
2 + 1 dimensional strongly coupled systems such as can occur at interfaces in condensed
matter experiments. In d = 4 the charged black hole has metric and timelike component of
38
the gauge potential At
ds2 =r2
L2(−fdt2 + d~x2) +
L2
r2
dr2
f, (10.1)
f(r) = 1 +Q2
r4− M
r3, At = µ
(1− r0
r
)(10.2)
where [ CHECK ESPECIALLY gF ]
µ =2Q
r0
(10.3)
The differences in the geometry as compared to the AdS-Schwarzschild solution Eq. (6.2)
are encapsulated by the altered redshift factor f(r). The horizon is still determined by the
largest root of this function, f(r0) = 0. We have used the gauge freedom to shift the gauge
field such that it vanishes at the horizon.
Using the recipe for the black hole thermodynamics of section ??, the black hole temper-
ature and entropy are
T =3r0
4πL2
(1− Q2
3r40
), s =
2π
κ2
(r0
L
)2
µ = 2gEMQ
r0
(10.4)
where r0 is related to M as
M = r30 +
Q2
r0
. (10.5)
The novel part of the charged black hole compared to the uncharged one is that one
notices that for r0 =√Q
4√
3the temperature vanishes, but the redshift factor stays non-trivial
T = 0⇔ r0 =
√Q
4√
3(10.6)
f(z) = (1− r0
r)2(1 + 2
(r
r0
)2
+ 3
(r
r0
)3
) (10.7)
One still has a black hole. This extremal black hole is the dual of a zero temperature finite
density state.
We can already make a remarkable statement about the state in the CFT that this ex-
tremal black hole encodes. Note that the entropy remains finite even for T = 0. This means
that the extremal AdS RN black hole encodes a groundstate with an extensive amount of
degeneracy. This is very counterintuitive, and strongly suggests that this system is actually
ridiculously unstable to any deformation that lifts the degeneracy.
39
10.1. AdS2 near horizon geometry and emergent local quantum criticality
Let us study the extremal black hole some more. In particular, just as the case for the
generic black hole, the characteristic physics is controlled by the horizon. In general, the
horizon rh is determined from the vanishing of f(rh) = 0. To study what happens near the
horizon we Taylor expand f(r) near r − rh as
f(r) = f ′(rh)(r − rh) + f ′′(rh)(r − rh)2 + . . . (10.8)
where the ‘. . . ’ are higher order terms in r − rh. Following AdS/CFT it is precisely this
near-horizon geometry that is representative for the IR physics in the field theory. In the
case of a generic black hole, we learned that f ′(rh) ∝ T . The finite temperature near horizon
geometry is therefore quite different from the zero temperature near horizon geometry. It
is governed by the next order coefficient f ′′(rh). This reflects the universal feature that
extremal black holes have a double zero at the horizon. Specifically
f(r) = 6(r − r0)2
r20
+ . . . (10.9)
Inserting the near-horizon redshift factor in the full metric yields the near horizon geom-
etry in the regime r−r0r0 1,
ds2 = −6(r − r0)2dt2
L2+
L2dr2
6(r − r0)2+r2
0
L2dx2 + . . . , (10.10)
while the gauge potential becomes
At =µ
r0
(r − r0) (10.11)
Something surprising has happened! One infers that in terms of the near horizon coordinate
r−r0 , the metric factors multiplying dt2 and dr2 acquire a similar structure as the bare AdS
metric Eq. (??): this is an effective anti-de Sitter geometry, except that the space directions
(dx2) are multiplied by the constant (r0)2! Therefore, the space directions form just a flat
space, while the effective anti de Sitter geometry is only realized in the two-dimensional t, r
directions! To render this more explicit, let us re-parametrize the near-horizon metric in
terms of a radial coordinate ζ which is the inverse of the distance from the horizon, and a
radius L2
ζ =L2
2
r − r0
, L2 =L√6, (10.12)
ds2 =L2
2
ζ2(−dt2 + dζ2) +
r20
L2d~x2, At =
1√6ζ
(10.13)
40
This is the metric of a space time with an AdS2 × R2 geometry, where the AdS2 part has a
rescaled radius L2 = L/√
6 w.r.t. the original AdS4 geometry.
The remarkable aspect of this near-horizon discovery is that the AdS2 geometry precisely
should encode a physical regime where time-like rescalings show self-similar behavior, but
spatial rescalings can be sensitive to an underlying fundamental scale. This is precisely what
the notion of local quantum criticality tries to capture. So the extremal AdS RN black hole
is the first description of a theory which contains in it in a very concrete and quantitative
way the notion of local quantum criticality. In the next lecture we will see how this manifests
itself in more detail.
11. EXERCISE: COMPUTING THE CONDUCTIVITY OF THE ADS-RN METAL
Following our introduction one of the characteristic quantities to compute in a finite
density system is the conductivity. Let us actually first compute the conductivity in the
µ = 0 AdS background. In the specific case of d = 2 + 1 dimensions, there is no need
for a detailed computation. In d dimensions a conserved current has dimensions d − 1.
Its Fourier transform therefore has dimensions −1. The two-point function without the
momentum-conserving delta function therefore has dimension d − 2, and the conductivity
σ(ω) = −1iω〈J(ω)J(ω)〉 thus has dimension d−3. In our case σ(ω) is therefore dimensionless.
Now in the pure AdS background dual to an exact CFT there is no scale in the problem.
Therefore the conductivity can only be a pure number independent of ω.
What happens when we turn on a chemical potential? In the presence of a chemical
potential, we now have an abundance of charge carriers at scales below µ. In the presence
of a constant electric field this will cause a current to run. In fact if we change nothing
else, this DC current will be infinite as the carriers start to accelerate. Combining this
with a sum-rule, that the integrated density of states (as a function of ω) cannot change
(this is unitarity/completeness of the underlying quantum theory), we see that between the
DC current (ω = 0) and the chemical potential-scale ω ∼ µ, the conductivity (which is
proportional to the number of states) must decrease compared to the µ = 0 conductivity.
As an exercise we will now see that this indeed happens in the conductivity of a finite density
CFT system through a computation of its AdS dual.
We will only work at T = 0 and for computational convenience, we will use Poincare
41
like coordinates for the isotropic extremal charged AdS black hole. In such coordinates the
background is
ds2 = gµνdxµdxν + gzzdz
2
=L2µ2
12z2(−f(z)dt2 + dx2 + dy2 +
12dz2
µ2f(z)) (11.1)
The metric interpolates between an asympotically AdS space at z = 0 and an extremal
horizon at the double zero of the redshift function f(z)
f(z) = (1− z)2(1 + 2z + 3z2) (11.2)
(Let us just mention for completeness that for the non-extremal finite temperature solution
f(z) has a single zero of course 7
f(z) = (1− z)(1 + z + z2 − q2z3), T =µ
8qπ(3− q2) (11.3)
and one should replace µ2/3 with µ2/q2 in the metric.)
Together with an electrostatic potential
A0 ≡ Φ = µ(1− z) (11.4)
it is a solution to the equations of motion of the AdS-Einstein-Maxwell action
S =1
2κ2
∫d4x√−g(R +
6
L2− L2
4FMNF
MN
)(11.5)
To obtain the optical conductivity at low-ω we will not use the Kubo formula, but Ohm’s law
directly. Recall that the conductivity is defined as J i = σEi. Recall also that in AdS/CFT
the dictionary computes the partition function in the presence of a source
Z(Jµ)CFT =
∫DXeiSCFT [X]+JµOCFT (X) (11.6)
Thus to study the CFT in the presence of an external (electric) field, Eiext = ∂tA
iext (in
the gauge A0ext = 0) we need that J i ∼ Aiext is non-vanishing. Through the dictionary
J iAdS/CFT≡ limz→0A
iAdS this means that leading source term to the equation of motion of
AiAdS is non-vanishing. At the same time the response J iCFTAdS/CFT≡ limz→0 ∂zA
iAdS (Here
we use that we know the scaling dimension ∆ = d− 1 for the operator dual to Ai). Putting
7 For our reference: compared to Sean’s lectures [1], we have set γ = 2, r+ = 2q/µ and rescaled r = r+z
42
this together we see that the conductivity just follows from solving the equation of motion
for AiAdS in the background and then evaluating the quantity
σ(ω) =JxCFTExCFT
= limz→0
∂zAxAdS(ω)
−iωAxAdS(ω)(11.7)
There is a tricky part to the computation. We need to evaluate the equation of motion
for fluctuations Ax in the charged extremal black-hole background. As we will see, when we
expand Aµ = ABHµ + δAµ and gµν = gBHµν + hµν the kinetic term will be off-diagonal in the
fluctuations δAµ, hµν . We will have to take this cross-coupling into account. So we take the
action (11.5) and expand to second order
S = SBH + δS|BH + δ2S|BH + . . . (11.8)
The first variation will vanish because the extremal BH is a solution to the equations of
motion. For the second variation, let’s consider the variation of the Maxwell term first. The
first variation gives the stress-tensor for the metric variation and the Maxwell equation for
the δAµ and then we need to vary this one more time
δ2S = δ
∫ √−g(−1
2hMNTMaxwell
MN − L2∂MδANFMN + . . .
)(11.9)
By symmetry δAx cannot mix with any of the other δAM fluctuations, so we can set all those
to vanish. Then from an expansion of the Maxwell stress tensor to first order in gauge-field
fluctuations Ax around the background
TMaxwellMN = gRSFMRFNS −
1
4gMNFRSF
RS
=(gRS(∂[MδAR])F
(BH)NS + (M ↔ N)
)− gMN(∂RδAS)FRS
(BH) + . . . (11.10)
=(gxx(∂MδAx)F
(BH)Nx − δMxg
RS∂RδAxF(bg)NS + (M ↔ N)
)− gMN(∂RδAx)F
Rx(BH) + . . .(11.11)
= δMxδNtgzz∂zδAxΦ
′ − δMxδNzgtt∂tδAxΦ
′ + (M ↔ N) + . . . (11.12)
In the last line we substituted that the only non-vanishing part of F(BH)MN = Fz0 = −F0z =
∂zABH0 ≡ Φ′. Thus we see that the Ax mixes with the metric fluctuations hxt and hxz.
Now we need to compute δ2SEinstein−Hilbert to compute their contribution. To do so, we
use a trick. Again, because of the symmetries all components of hMN where neither M = x
or N = x can be set to zero consistently because of the symmetries of the system. Also
we are only interested in frequency-dependent fluctuations and not momentum-dependent
43
ones. Now, the trick is that an off-diagonal metric fluctuation with one component in a
spatial direction hxM(z, t) that does not depend on the x-direction behaves effectively as a
gauge-field with a spacetime dependent coupling [41, 52]
SEHeff =
∫d4x√−gbg
(− 1
4geff(z)2f
(x)MNf
(x)MN
)(11.13)
with g2eff = gxx, A(x)
M = hxM and gbg the background metric.
Analyzing this part by part, by symmetries again Ay decouples anc can be to zero. Next,
the equation of motion for Az is a constraint, and directly enforcing this constraint into
the action the remaining two modes At and Ax decouple. To see this, note that for zero
momentum, ~k = 0, the only nonzero component of fMN is fzt. Fourier transforming all fields
in the time-direction φ(t) =∫
dω2πφ(ω)e−iωt, the effective action therefore equals (as usual
quadratic terms a2z below should be read as az(−ω)az(ω))
Seff = δ2SEHeff + Sinteff
=
∫ √−gbg
(− 1
2gxxgzzgtt
(∂zA2
t + ω2A2z + 2iω(∂zAt)Az
)− 1
2gzzgttAtΦ′∂zAx +
iω
2gzzgttAxAzΦ′
)(11.14)
The second term Sinteff is the first stress term obtained in eq. (11.9), where for simplicity Ax
is what we called δAx before.
The constraint that follows from varying w.r.t. Az(ω) equals
− ω2Az − iω∂zAt +iω
2gxxAxΦ
′ = 0 (11.15)
Substituting the constraint back into the total action, including the second term in (11.9),
one finds
S =
∫ √−gbg
(− 1
gxxgzzgtt
((∂zAt)2
))+
1
2gzzgtt∂zAtAxΦ′ −
1
8gzzgttgxxA2
x(Φ′)2
− 1
2gzzgttΦ′(∂zAx)At
−gxxgzz
2(∂zAx)
2 − ω2gxxgtt
2A2x
)(11.16)
Integrating the term in the second line by parts and using the background equation of motion
∂z√−ggzzgttΦ′ = 0 (11.17)
44
one finds the decoupled system (after shifting ∂zAt)
S =
∫ √−gbg
(− 1
gxxgzzgtt
((∂zAt − gxxAxΦ′)2
))+
1
8gzzgtt(Φ′Ax)
2gxx
−gxxgzz
2(∂zAx)
2 − ω2gxxgtt
2A2x
)(11.18)
Using eq. (7.24) and eq. (5.12) we can now directly write down the equation of motion for
Ax equals
1√−g
∂z√−ggxxgzz∂zAx − ω2gxxgttAx +
1
2gttgxxgzz(Φ′)2Ax = 0 (11.19)
Substituting the explicit metric of the extremal AdS-RN black hole, this simplifies to8
∂zf(z)∂zAx +4q2ω2
µ2f(z)Ax −
4q2z2
2µ2(Φ′)2Ax = 0 (11.20)
The final step is to solve this equation numerically, and extract the objective of interest
σ = limz→0
i
ω
∂zAxAx
(11.21)
We will do so with Mathematica in the exercise session. What one finds (Fig 4), is indeed the
expected behavior of the conductivity, with an infinite peak at ω = 0 (by Kramers-Kroenig
this is 1/ω in the imaginary part), a dip for ω < µ and a constant asymptote for ω µ.
FIG. 4: The conductivity as a function of ω/T . When ω → 0, the imaginary of the optical
conductivity behaves as 1/ω. This reflects the existence of a delta-function δ(ω) in Reσ. In these
plots, from top down, the chemical potential grows larger. These plots are taken from [1]
8 This equals [CHECK: There is a factor 2 difference in the Φ′ term. This can be easily ’fixed’
by adjusting the normalization of the stress-tensor.... It should be the factor of 2 just found
in eqn (11.18)] eqn (110) in [1] after the identification z = µ2q r = r/r+, r+ = 2q/µ and γ = 2.
45
0 2000 4000 60000
20
40
60
80
7 K 100 K 160 K 200 K 260 Kar
ctan
(VV
) (d
egre
e)
Wavenumber (cm-1)
1
10
100 1000
b
a
100 K 160 K 200 K 260 K
|VZ| = c Z-0.65
Wavenumber (cm-1)|VZ|
(kS
/cm
)
Figure 3 Universal powerlaw of the optical conductivity and the phase angle spectra of optimally dopedBi2Sr2Ca0.92Y0.08Cu2O8+į. The sample is the same as in Fig. 1. In a the phase function of the opticalconductivity, Arg(ı(Ȧ)) is presented. In b the absolute value of the optical conductivity is plotted on a doublelogarithmic scale. The open symbols correspond to the powerlaw |ı(Ȧ)|=CȦ–0.65.
FIG. 5: Left: Observed power law |σ(ω)| = Bω2/3 in the optical conductivity measured in the strange
metal phase of high Tc superconductors at optimal doping Experiments [73]. Right: Observed
power law |σ(ω)| = Bω2/3 + C in the optical conductivity of a strongly coupled system in the
presence of an ionic lattice computed numerically through AdS/CFT [72].
11.1. The promise of AdS/CFT
This very straightforward computation resonates in an extremely tantalizing current de-
velopment. One of the mysteries of high-Tc superconductors is that the optical (= frequency
dependence of the) conductivity in the strange metal phase shows an intermediate scaling
behavior (Leftpanel in Fig. 5) with a power σ ∼ (iω)−2/3 [73]. The remarkable finding is
that an AdS/CFT computation in the background of a “modulated” RN black hole with
a chemical potential µ = µ0 + µ1 cos(kx) to account for underlying atomic crystal lattice
effects, has an almost identical ∼ (iω)−2/3 scaling regime [72]! This is super-tantalizing, and
people are actively trying to understand if there is indeed a relation between this intricate
numerical computation and experiment and what it is.
46
Lecture IV
12. HOLOGRAPHIC NON-FERMI LIQUIDS
We have seen that with a lattice perturbation the conductivity of the extremal AdS-RN
black hole bears a remarkable resemblence to that measured in the strange metal phase of
high Tc superconductors. We will now try to probe the strange metal more directly. Recall
from our review of the Fermi liquid theory that a successful phenomenological model of
the strange metal is the marginal Fermi liquid. This a special type of non-Fermi-liquid
characterized by a self-energy Σ(ω) ∼ ω lnω in the spectral function.
The spectral function is equivalent to the imaginary part of the retarded Green’s function
of a fermionic excitation. But (two-point) correlation functions is precisely what AdS/CFT
is good at calculating. Let us therefore compute the spectral function of a fermionic operator
in a strongly coupled CFT at finite density dual to the AdS-RN black hole.
12.1. Fermionic correlation functions from holography
The novelty compared to our previous computations is the spin 1/2 nature of a fermionic
operator. Following the dictionary this is dual to a spin 1/2 field, and to ensure that it is the
field whose condensed matter aspects we wish to study we make it charged. The minimal
AdS action that describes the gravity dual of this system is the AdS-Einstein-Maxwell-Dirac
action
S =
∫dD+1x
√−g(R− 2Λ− 1
4FµνF
µν − Ψ[eµaΓa(∂µ +
1
4ωµabΓ
ab − iqAµ)−m]Ψ
).
(12.1)
Because fermions transform non-trivially under the Lorentz group, they also couple in a
special way to the spacetime background. This is encoded through both the coupling with
a vielbein, the “square root” of the metric defined by eaµebνηab = gµν with ηab the fixed
Minkowski metric, and a spin connection ωµab defined through the demand that the vielbein
should be covariantly constant
Dµeaν = 0 = ∂µe
aν − Γσµνe
aσ + ωaµbe
bν . (12.2)
47
This machinery ensures that the matrices Γa are the usual fixed Dirac-matrices obeying the
anti-commutation relation Γa,Γb = 2ηab.
Due to the first order nature of the Dirac action, there are some technical changes to the
prescription of applying the AdS/CFT dictionary. That such changes are necessary is most
readily apparent by counting components of fermions. In dimensions d = 2n and d = 2n+ 1
a spinor has 2n components. This means that if one has an even dimensional bulk with
d = 2n the spinor has double the components, 2n than one would naively expect based on
the dimensionality d = 2n− 1 of boundary. Due to the first order nature of the action this
counting is wrong. The first order action simultaneously describes the fluctuation — half of
the components — and its conjugate momentum — the other half. Clearly only the former
should correspond to a boundary degree of freedom. The most ready way to do so is to use
the extra direction to project the fermion into two distinct eigenstates Ψ± of Γr and call
one the fluctuation, say Ψ+. The other Ψ− is then the conjugate momentum.9 Under this
projection the Dirac action reduces to
S =
∫d4x√−g(−Ψ+/DΨ+ − Ψ−/DΨ− −mΨ+Ψ− −mΨ−Ψ+
). (12.3)
The second issue is that the AdS/CFT correspondence instructs us to derive the CFT
correlation functions from the on-shell action. The Dirac action, however, is proportional
to its equation of motion. This reflects the inherent quantum nature of fermions.The action
therefore appears to vanish on-shell. Recall, however, that in the case of scalars the full
contribution came from a boundary term, and we did not specify this yet. In the case
of the scalar the boundary term appeared naturally. Here we have to do some work to
determine it. Having chosen Ψ+ as the fundamental degree of freedom, we will choose
a boundary source Ψ0+ = limr→∞Ψ+(r). Then, because the dynamical equation is first
order, the boundary value Ψ0− is not independent but related to that of Ψ0
+ by the Dirac
equation. We should therefore not include it as an independent degree of freedom when
taking functional derivatives with respect to the source Ψ0+. Instead it should be varied to
minimize the action. To ensure a well-defined variational system for Ψ− we add a boundary
9 Due to the first order nature of the action a bulk Dirac spinor in d = 2n dimensions with 2n−1 (complex)
degrees of freedom corresponds to 2n−1 component complex Dirac operator on the d = 2n−1 dimensional
boundary. A bulk Dirac spinor in d = 2n+1 dimensions with 2n−1 complex degrees of freedom corresponds
to a 2n−1 component Weyl spinor on the d = 2n dimensional boundary.
48
action,
Sbdy =
∫r=r0
d3x√−h Ψ+Ψ− (12.4)
with hµν the induced metric. The variation of δΨ− from the boundary action,
δSbdy =
∫r=r0
d3x√−h Ψ+δΨ−
∣∣∣∣Ψ0
+fixed
, (12.5)
now cancels the boundary term from variation of the bulk Dirac action
δSbulk =
∫ √−g(−δΨ+(/D)Ψ+ − ((/D)Ψ+)δΨ+ − δΨ−(/D)Ψ− − ((/D)Ψ−)δΨ−
)+
∫z=z0
√−h(−Ψ+δΨ− − Ψ−δΨ+
)∣∣∣∣Ψ0
+fixed
. (12.6)
The next complication is that the fermionic correlation function is in general a matrix
between the various spin components. A completely covariant formulation exists [100, 123],
but with some insight we can separate out each independent spin component [101]. This is
very similar to a standard analysis of the Dirac equation in flat space as in any standard
QFT course. Undoing the projection unto Γr, we first write the full covariant first order
Dirac equation (Γaeµa(∂µ +
1
4ωµabΓ
ab − iqAµ)−m)
Ψ = 0 (12.7)
Note that the boundary term does not contribute. The first step is the insight that metrics
whose components depend on only a single coordinate, such as AdS-black holes, have the
property that by a redefinition of the Dirac fields,
Ψ = (−ggrr)−1/4ψ (12.8)
the spin connection ωµab can be removed. This simplifies the Dirac equation to the more
standard form
(Γaeµa(∂µ − iqAµ)−m)ψ = 0 (12.9)
We shall also only be interested in background configurations with just A0 ≡ Φ(r) non-zero
and a function of the radial AdS direction r only.
In a flat spacetime we would now Fourier transform and project the 4-component Dirac
spinor onto 2-component spin-eigenstates. Since the radial direction of anti-de-Sitter space
49
breaks the four-dimensional Lorentz invariance, this cannot be done here. However, a similar
projection exists onto transverse helicities, where the spin is always orthogonal to both the
direction of the boundary momentum and the radial direction. We choose the basis for our
d = 3 + 1 dimensional Dirac matrices
Γr =
−σ311 0
0 −σ311
, Γt =
iσ111 0
0 iσ111
, Γx =
−σ211 0
0 σ211
, Γy =
0 −σ211
−σ211 0
(12.10)
and we Fourier transform only along the directions t, x, y parallel to the boundary
ψ(r, t, x, y) =∫
dω2π
dkx2π
dky2πeikxx+ikyy−iωtψ(r, ω, kx, ky). Using rotational invariance we are free
to choose the boundary momentum along the x-direction, ~k = (kx, 0). With this choice for
the momentum one can show that the operator Γ5Γy commutes with the Dirac operator.
We can therefore project onto its eigenstates χ1,2. This reduces the Dirac equation to an
equation for the two-component t-helicities
√grr (−σ3∂r −
√grrm)χi(r;ω, k) = −(iσ2
√gxxkx + σ1
√−gtt(ω + qΦ))χi (12.11)
Here we have use a shorthand that for a diagonal metric gµν |µ6=ν = 0, the vielbein emµ is
literally the square root emµ =√gµµδ
mµ .
It suffices to consider only χ1 from here on, as the results for χ2 simply follow by changing
kx → −kx. [I AM NOT SURE THIS IS CORRECT]. To construct the CFT correlation
function the essential part is the behavior of the solution near the AdS boundary. In that
region√grr = r/L+ . . . ,
√gxx =
√gtt = L/r + . . . thus to leading order in r →∞ the eqn
(12.11) behaves as (∂r +
L
rσ3m
)χ1(r;ω, k) = 0 (12.12)
Similar to scalar case near the AdS boundary the 2-component spinor χ1 has the asymptotic
behavior
χ1(r) = a
0
1
rmL + b
1
0
r−mL + . . . (12.13)
For a generic mass the first component is non-normalizable, but the second one is.
By acting with the operator(∂r − iσ2
√grrσ1m
)(−iσ2)
√grrgii
on the equation for χ1 one
can obtain a second order equation for each of the individual components of χ1, with two
50
independent solutions. Given these homogenous solutions to the Dirac equation, the (bulk)
Green’s function for χ+, still a two by two matrix, can be constructed. It equals
G(ω, k, r1, r2) =ψb(r)⊗ ψint(r′)θ(r − r′)− ψint(r)⊗ ψb(r′)θ(r′ − r)
12
(ψint(r)σ3ψb − ψb(r)σ3ψint
) . (12.14)
Just as for the scalar case, ψb(r) is the normalizable solution with leading coefficient a = 0.
and ψint(r) is determined by the appropriate boundary conditions in the interior.
We can now use this bulk-Green’s function to evaluate the on-shell Dirac action from
which we can find the boundary correlation function. The on-shell Dirac action comes fully
from the boundary term, we argued should be there in eqn (12.4). There we postulated
that the fundamental field is Ψ+ ≡ 12(1 + Γr)Ψ. For the t-helicity χ1 this reduces to the
projection with respect to 12(1− σ3).
At the same time one should think of Ψ− as dependent on the fundamental field Ψ+.
Since limr→∞Ψ+(r) is the source, one can find the dependence of Ψ− from the bulk Green’s
function
Ψ−(r) = limr2→∞
∮1
2(1 + σ3)G(r, r2)Ψ+(r2) (12.15)
Substituting this Green’s function into the boundary action one obtains after projection on
the independent t-helicity χ1 [MINUS SIGN MISSING]
S = limr→∞
∫ √−hχ0
1
1
2(1 + σ3)
ψint(r)⊗ ψb(∞)12
(ψint(r)σ3ψb − ψb(r)σ3ψint
)χ01. (12.16)
In the t-helicity basis the boundary source (the non-normalizable coefficient) has only a
lower component
χ01 =
0
J
. (12.17)
Writing also ψint(r) =
bintr−m + . . .
aintrm + . . .
, ψb(r) =
br−m + . . .
0 + . . .
, one finds
S = limr→∞
∫ √−hr−2m (J†bint)(b
†J)
b†aint + a†intb. (12.18)
The final step is that an inspection of the Dirac equation reveals that at zero density in pure
AdS one can always choose b and aint real (but not bint).
S = limr→∞
∫ √−hr−2mJ†
bintaint
J. (12.19)
51
Differentiating the on-shell action w.r.t J and J†, and dropping both the int subscript
and the overall r−2m term gives the expression for the fermionic CFT correlation function
Gfermions =b
a. (12.20)
12.1.1. On various Green’s functions...
We mention now a subtlety that we brushed over earlier in the lectures. Recall that in
general there are various types of Green’s functions, Feynman, advanced, retarded etc. In
principle which one you wish is determined by the apporpriate boundary conditions in the
interior. However, note that the Green’s function obtained by differentiating the boundary
term in the action is always real (because the action is always real). Something appears
amiss, because the advanced and retarded Green’s functions are generically complex. A
very careful finite temperature field theory analysis of AdS/CFT reveals that one in fact
simply has to take the result (12.20) in all cases even when b and a are not manifestly real.
12.2. Non-Fermi liquids from holography
To study the spectral function of fermionic operators in the background of the holographic
AdS-RN metal, we must therefore solve the Dirac equation (12.11) in the background of
the extremal RN black hole and take the ratio of the subleading to leading fall-off near the
AdS boundary.
A beautiful paper in the summer of 2009 showed how to do this semi-analytically using
a standard trick from quantum mechanics: matching the asymptotic solutions near the
boundary to the asymptotic solution near the horizon in an overlapping range of applicability.
We already know what the solution looks like near the AdS boundary, see Eq. (12.13). Near
the horizon r − r0 1 the metric is given by the AdS2 × R2 metric Eq. (10.10), and thus
the Dirac equation becomes√
6
L(r − r0)(−σ3)∂rχ1 = −
(iσ2
L
r0
kx −m−L√
6(r − r0)σ1(ω + q
µ
r0
(r − r0))
)χ1 (12.21)
Changing coordinates to ρ = ωL6(r−r0)
this equals
ρ∂ρχ1 = −L(σ1
kxL√6r0
− iσ2ρ− iσ2qµL
6r0
− σ3m√
6
)χ1 (12.22)
52
Acting with ρ∂ρ on both sides and using the Dirac equation, this equals the second order
equation
ρ2∂2ρχ1 + ρ∂ρχ1 = L2
(σ1
kxL√6r0
− iσ2ρ− iσ2qµL
6r0
− σ3m√
6
)2
χ1 + iσ2Lρχ1 (12.23)
=
(k2xL
4
6r20
+m2L2
6− q2µ2L2
36r20
)χ1 + iσ2Lρχ1 (12.24)
The region ρ 1 is now the near-horizon region, whereas the region ρ 1 is the boundary
of the AdS2 region where the near-horizon metric starts to break-down. This is where we
should match to the asymptotic solution from the AdS4 boundary. In this transition region,
the boundary of AdS2, the equation becomes diagonal
ρ2∂2ρχ1 + ρ∂ρχ1 =
(k2xL
4
6r20
+m2L2
6− q2µ2L2
36r20
)χ1 + . . . (12.25)
and can be solved in terms of a simple power of ρ
χ ∼ αρ−νk + βρνk + . . . (12.26)
where νk =√
k2xL4
6r20+ m2L2
6− q2µ2L2
36r20. We are in particular interested in the ω-dependence of
the solution. This can be made manifest by defining ρ = ωζ.
χ ∼ αω−νkζ−νk + βωνkζνk + . . . (12.27)
Note, however, that ζ is literally proportional to the natural coordinate on AdS2. Thus
following the definition of the Green’s function as the ration of the subleading to the leading
fall-off, the combination βω∆AdS2/α is precisely the AdS2 Green’s function. Rescaling the
solution, it can therefore be written in a very suggestive way as
χ1 ∼ η+ζ−νk + GAdS2(ω)η−ζ
νk + . . . (12.28)
with η± a pair of ω-independent spinor s, G(ω)AdS2 ∼ ω2νk .
This asymptotic expansion is valid in the regime where r − r0 . r0. The asymptotic
expansion from the boundary on the other hand is valid roughly for any r ω. Thus
there is an overlapping regime whenever ω < 2r0. For such small ω we can match at any
point in the overlapping region. Generically each independent solution of the one expansion
will be a linear combination of the independent solutions of the other expansion. Thus the
asymptotic boundary behavior abstractly is of the form
a = a+(ω) + a−G(ω) (12.29)
b = b+(ω) + b−G(ω) (12.30)
53
where a± corresponds to the coefficient of the near-boundary solutions. This then allows
us to express the (retarded) Green’s function of the CFT for small ω in terms of the near
boundary solutions
GR(ω) =b+(ω) + b−G(ω)
a+(ω) + a−G(ω)(12.31)
In particular near ω = 0 it will behave as
GR(ω, k) =b
(0)+ + ωb
(1)+ +O(ω2) + Gk(ω)
(b
(0)− + ωb
(1)− +O(ω2)
)a
(0)+ + ωa
(1)+ +O(ω2) + Gk(ω)
(a
(0)− + ωa
(1)− +O(ω2)
) . (12.32)
This is a remarkable result, because even though we do not know the exact expressions for
a(i)± (k) and b
(i)± (k), we can already deduce the nature of the Fermi-liquid quasiparticles, if
there are any. The assumption that there are such particles is equal to the assumption that
the Green’s function has a pole at ω = 0. This occurs if a(0)+ (k) vanishes for a specific kF .
Indeed near the Fermi surface, the retarded Green’s function becomes of the canonical form
GR(ω, k) ' Z
ω − vF (k − kF )− Σ(k, ω)+O(k − kF , ω), , (12.33)
with the self-energy Σ(k, ω) = G(ω)
(a(−0)
a(1)+
+O(ω) The one aspect we do know already is the
frequency dependence of the self-energy. It is Σ ∼ ω2νkF . But this scaling is precisely what
we wish to know, because the defining property of a Landau Fermi liquid is that Σ ∼ ω2.
Our calculation thus indeed shows that generically the fermionic excitations in the AdS-RN
metal are non-Fermi liquids.
The precise details of the Fermi liquid — and whether there is indeed such a pole —
depend sensitively on the parameters of the theory as νk depends directly on the charge and
scaling dimension of the Fermionic operator. Explicit numerical calculation has shown that
generically for large enough charge compared to the scaling dimension such a pole exists
(Figs ?? and 6) with the specific properties listed in table II.
13. HOLOGRAPHIC SUPERCONDUCTORS
The CFT state dual to the AdS-RN black hole thus has remarkable properties, that for
a specific choice of parameters can tantalizingly mimick the observed spectral functions in
the strange metal phase of high Tc superconductors. The pure AdS-RN black hole has one
54
0.5 1.0 1.5 2.00
0.02
0.04
0.06
0.08
!/|µ0|
A (
!,
k)
k/T = 0
k/T = 1.5
k/T = 1.9
k/T = 2.3
!2"−d fit
0!0.5 0.5 1.00
0.01
0.02
0.03
0.04
!/|µ0|
A (
!,
k)
k/T = 0
k/T = 1.5 Teff
/T
k/T = 1.9 Teff
/T
k/T = 2.3 Teff
/T
k < kF
(!!|µ0|)
2"!d tail
k > kF
1.7 3.4! 1.7 0!/T
eff
k = 0
EF = ! 0.19
FIG. 6: Left: Emergence of a Fermi surface from a critical point as µ/T is increased [100] Right:
The false color “ARPES” pictures of T/µ 1 holographic Fermi liquids: Top, νkF < 1/2, non-
Fermi liquid; Middle, νkF = 1/2, marginal Fermi liquid; Bottom, νkF > 1/2, irregular Fermi liquid.
Plots are taken from [102].
Σ ∼ ω2νkF Fermi-system Phase Quasiparticle Properties: (dispersion, peak)
2νkF > 1 regular FL ω∗(k) = vFk⊥, Γ(k)ω∗(k) ∝ k
2νkF−1
⊥ → 0, Z = constant
2νkF = 1 marginal FL GR = h1c2ω−vF k⊥+cRω lnω , Z ∼ 1
lnω∗→ 0
2νkF < 1 singular FL ω∗(k) ∼ k1/2νk⊥
Γ(k)ω∗(k) → const, Z ∝ k
1−2νkF2νkF⊥ → 0
TABLE II: The zoo of non-Fermi liquids one can find in AdS/CFT. .
in-your-face problem, however: its extensive groundstate entropy. Strictly speaking the RN-
black hole therefore describes the typical state amongst a set of degenerate states that scales
as the volume of the system. This suggests that the system is in fact immensely unstable.
55
Let’s see if we can understand these instabilities. When we considered only the Einstein-
Maxwell sector in the AdS theory, we concentrated on only the energy-momentum and
electromagnetic current sector of the dual CFT but ignored all other excitations in the
theory. The most plausible explanation is that the instabilities are driven by these ignored
excitations.10 In the previous section we considered adding fermions to the AdS-Einstein-
Maxwell theory. For the purpose of instabilities, we’ll make it even simpler and consider an
additional charged scalar field. The action we consider is thus
S =
∫dD+1x
√−g(R− 2Λ− 1
4FµνF
µν − (∂µ + iqAµ)φ(∂µ − iqAµ)φ−m2φφ
). (13.1)
13.1. (In)stability analysis
Let us first perform a standard stability analysis and consider the field φ as a small
fluctuation around the extremal AdS-RN background. After Fourier transforming along the
boundary directions the equation of motion for φ reads
1√−g
∂r√ggrr∂rφ+ gtt(ω + qAt)
2 − giik2 −m2L2φ = 0 (13.2)
Substituting the background for the extremal AdS4-RN black hole, we have
1
r2∂rr
4f(r)∂r +1
r2f(r)(ω + qAt)
2 − 1
r2(k2)−m2L2φ = 0 (13.3)
Since we are considering φ as a fluctuation, and since any instability should require very
little energy, i.e. we can concentrate on small ω fluctuations, we can perform the same
matching analysis as for the fermions to study the solutions to the equation of motion.
Near the AdS-boundary we had φ = Ar∆−d +Br−∆ + . . . with
∆ =3
2+
√9
4+m2L2 (13.4)
Now note that the scaling dimension stays real for all values m2L2 > −9/4. It appears the
equation makes sense for negative m2. This sounds strange, because normally we associated
a negative m2 of a scalar fluctuation as a tachyonic instability where the scalar sits at the
top of a local maximum of the potential. What happens is that in AdS is that there is an
10 The precise way of saying this is that the Einstein-Maxwell theory is not a complete quantum theory of
gravity. Other fields are necessary to have a completely consistent theory.
56
additional gravitational contribution to the potential which modifies the tachyonic instability
precisely to the location where ∆ turns imaginary. This is easily seen by taking the flat space
limit L 1 which connects the stability boundary ∆ = 0 directly to m2 = 0. So in pure
AdS we may consider slightly negative scalar m2 and still have a stable system.
In the AdS2-region (10.12) on the other hand the equation becomes
ζ2∂2ζφ+ ζ2(ω +
q
6ζ)2φ− k2
r20
−m2L22φ = 0 (13.5)
Its asymptotic behavior near the boundary is
φ ∼ αζ1−∆AdS2 + βζ∆AdS2 + . . . (13.6)
where ∆AdS2 = 12
+ νk with νk =√
k2
r0
2+m2L2
2 −q2
6. From these we read off that near the
horizon in the AdS2 region the scalar is unstable when m2L22 = m2L2/6 > −1
4. We have
indeed found an instability. A zero-momentum scalar mode therefore has a mass-window
−6
4 m2L2 −9
4. (13.7)
where the system has a tachyonic instability near the horizon, but well-defined as a theory.
The next question is what the new grounstate is that arises from this instability.
13.2. The holographic superconductor solution
Before we construct this new groundstate, let us discuss its physics. The resulting solution
will be a gravitational system with a non-zero electric field partially if not fully sourced by a
non-vanishing scalar profile. Since this profile must extend to the AdS boundary, there is a
directly translatable consequence in the CFT. The groundstate has a nonvanishing charged
expectation value for the scalar field, i.e. the U(1) symmetry is spontaneously broken. A
more precise way of saying it is that we have Higgsed the dynamical gauge field in the bulk
and this corresponds to global symmetry breaking on the dual boundary. It is in this Landau-
Ginzburg sense (where the U(1) of electromagnetism is considered a global symmetry) that
the new groundstate is a superconductor.
The full holographic superconductor solution involves solving the coupled AdS-Einstein-
57
Maxwell-Scalar system
Rµν −1
2gµν(R− 2Λ) = κ(TEMµν + T scalarµν ) (13.8)
DµFµν = iq(φ∂µφ− φ∂µφ) + 2q2Aµφφ (13.9)
1√g
(∂µ − iqAµ)gµν√−g(∂ν − iqAν)φ−m2φ = 0 (13.10)
One can solve the whole system, see e.g. [78, 129, 130], but the essential physics is already
captured in a simple limit. The dynamics of interest is in the U(1) sector and to emphasize
this we can take the limit q κ/L2. Since we want to keep the dynamics of the scalar field,
this means we should do so after rescaling Aµ → Aµ/q. We also wish to keep the Aµφφ
interaction, so we also first rescale φ → φ/q. After this rescalings the stress-energy tensor
becomes Tµν → 1qTµν , so in the limit q →∞, the dynamics decouples from the gravity sector
completely. One has the effective equations of motion (dropping the hatted notation)
Rµν −1
2gµν(R− 2Λ) = 0 (13.11)
DµFµν = 2Aµφφ (13.12)
1√−g
(∂µ − iAµ)gµν√−g(∂ν − iAν)φ−m2φ = 0 (13.13)
The solution to the vacuum AdS-Einstein equation is just the AdS-Schwarzschild black hole
6.2. We now make the ansatz that the solution will maintain the translational and rotational
invariance of the black hole. We posit
φ = φ(r) , A0 = Φ(r) , Aµ6=0 = 0 (13.14)
Moreover we can take φ(r) to be purely real without any loss. The Maxwell-scalar sector
then becomes
1
r2∂rr
2∂rΦ =2
f(r)r2Φφ2 (13.15)
1
r2∂rf(r)r4∂rφ−m2φ = − 1
f(r)r2Φ2φ (13.16)
The groundstate we seek is then the solution to these equations subject to the appropriate
boundary conditions. We wish to study the system at a finite chemical potential. Thus near
the AdS boundary Φ(r) = µ + . . .. Moreover, we do not want an explicit source for the
scalar field, thus it must fall-off near the boundary. With these two conditions, it is easy to
58
see that the near-AdS boundary behavior is exactly the same as for the fluctuations, with
the condition that we only consider the normalizable mode for φ.
Φ = µ− ρ
2r+ . . . (13.17)
φ = Or∆ + . . . (13.18)
We can therefore immediately interpret O as the gravitational encoding of the expectation
value of the CFT scalar operator dual to φ. It will be the order parameter in the dual theory.
Now µ is fixed, but the charge density ρ and the order parameter vev O are unknown.
The system is still underdetermined, since we need two more boundary conditions. These
are determined at the horizon of the black-hole. This is the location where the redshift
factor f(r) vanishes. Consistency of the Maxwell equation (13.17) then indicates that either
Φ or φ has to vanish at the horizon as well. If one chooses the latter, the only reasonable
solution will be φ = 0 everywhere. Therefore one boundary condition is that Φ(rh) = 0.
Substituting this in the scalar field equation one finds that φ(r) at the horizon must satisfy
∂rφ = m2φ/(f ′(rh) + 4/rh).
We now have a uniquely determined system and it remains to find the exact solution. In
practice the only way to do so is numerically. The subtle part is that one has to impose
boundary conditions on both boundaries. There are several numerical recipes to do so. The
most common is a shooting method, where one fixes two of the boundary conditions at the
horizon and varies the remaining two until one finds a solution that obeys the asymptotically
AdS-boundary conditions. With modern computers this can be done quite fast. One finds
that for high T/µ or high m/q there is no non-trivial solution. But for a q/m & 1 there is
a critical temperature/chemical potential ratio, for which a non-trivial solution exists; see
Fig. 7.
13.2.1. The macroscopic properties of the holographic superconductor
The characteristic feature of a superconductor is of course the vanishing of the resistance.
Using the tools we developed earlier in computing conductivities with AdS/CFT we can also
compute the conductivity of the holographic superconductor. The results, compared to the
RN black hole, are given in the right panel in Fig. 7. One notes in fact that for T < Tc
there is a remnant conductivity. However as this keeps decreasing as T is lowered one sees
59
0.0 0.2 0.4 0.6 0.8 1.00
2
4
6
8
10
T
Tc
qO2
Tc
0 10 20 30 40 50 60 700.0
0.2
0.4
0.6
0.8
1.0
1.2
ΩT
Re@ΣD
FIG. 7: Left: The condensate O2 as a function of temperature for q = 3,m2 = −2. Near the
critical temperature, O ∝ (Tc−T )1/2. [FIGURE SOURCE?] Right: Electrical conductivity for
holographic superconductors of O2 with q = 3,m2 = −2 (red line) and RN black holes (blue line):
T/Tc = 1, 0.775, 0.274 from top to bottom.The curves at T = Tc coincide. There is a delta function
at the origin in all cases. FIG. Source[IS THIS AN ORIGINAL FIGURE]
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-100
-50
0
50
100
150
200
TTc
FHVTc3L
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.40
100
200
300
400
TTc
SHVTc2L
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.40
100
200
300
400
500
600
700
TTc
CvTc2
FIG. 8: The free enengy, entropy and specific heat for the RN black hole (blue dashed line) and
hairy black hole (red solid line) of O2 as a function of temperature for q = 3,m2 = −2. It is easy
to see that below Tc the hairy black hole is a more stable state. FIGURE SOURCE
the resistivity shrinking. For any finite T there is always a tiny resistivity left.
Similarly we can compute other macroscopic properties of the holographic superconduc-
tor. They are shown in Fig. 8.
Even though the qualitative physics of the holographic superconductor should simply be
seen as standard spontaneous symmetry in the language of the dual gravitational theory, we
should remark that quantitatively the holographic superconductor differs from spontaneous
symmetry breaking in weakly coupled non-critical theories. Specifically in weakly coupled
quantum field theories the canonical scaling dimension ∆ of a scalar field is always infinites-
imally close to the free scaling dimension ∆free = (d − 2)/2. The scaling dimension of the
60
holographic superconductor, however, is a completely free parameter encoded in its AdS
mass m2L2 = ∆(∆− d). One consequence is that the dynamical onset of superconductivity
measured through its susceptibility differs [94].
Let us end with a brief observation about the revolutionary nature of the holographic
superconductor. Conventionally the viewpoint about black holes is that they are the most
stable objects around. Matter can only fall in, and classically never come out. The intuition
is that Because gravity is attractive, once you have a black hole, it will always be there.
Quantum mechnically the black hole will radiate, but this is not a strong instability of
the system. The astonishing aspect of the holographic superconductor solution is that it
shows that for AdS-black holes this wisdom does not hold. In AdS space black holes can
be unstable in the classical sense. The holographic superconductor is simply the first such
solution. By now we have legions of examples. There are p-wave and d-wave holographic
superconductors, where spin-1 and spin-2 fields condense; there are striped superconductors,
where the groundstate is not translationally invariant, etc.
14. EXERCISES
In the final exercise of this session we will discuss what happens if the instability is driven
by spin 1/2 fermions.
Exercise 1: Construct the electron star background and compute its macroscopics.
[1] S. A. Hartnoll, “Lectures on holographic methods for condensed matter physics,” Class.
Quant. Grav. 26, 224002 (2009) [arXiv:0903.3246 [hep-th]].
[2] C. P. Herzog, “Lectures on Holographic Superfluidity and Superconductivity,” J. Phys. A A
42, 343001 (2009) [arXiv:0904.1975 [hep-th]].
[3] J. McGreevy, “Holographic duality with a view toward many-body physics,” Adv. High
Energy Phys. 2010, 723105 (2010) [arXiv:0909.0518 [hep-th]].
[4] S. A. Hartnoll, “Horizons, holography and condensed matter,” arXiv:1106.4324 [hep-th].
[5] N. Iqbal, H. Liu and M. Mezei, “Lectures on holographic non-Fermi liquids and quantum
phase transitions,” arXiv:1110.3814 [hep-th].
61
[6] S. M. Carroll, “Spacetime and geometry: An introduction to general relativity,” San Fran-
cisco, USA: Addison-Wesley (2004) 513 p
[7] R. M. Wald, “General Relativity,” Chicago, Usa: Univ. Pr. ( 1984) 491p
[8] S. Weinberg, “Gravitation and Cosmology. Principles and Applications of The General The-
ory of Relativity,” John Wiley & Sons, 1972, New York.
[9] C. W. Misner, K. S. Thorne and J. A. Wheeler, “Gravitation,” San Francisco 1973.
[10] J. M. Maldacena, “The large N limit of superconformal field theories and supergravity,”
Adv. Theor. Math. Phys. 2, 231 (1998) [Int. J. Theor. Phys. 38, 1113 (1999)] [arXiv:hep-
th/9711200].
[11] O. Aharony, S. S. Gubser, J. M. Maldacena, H. Ooguri and Y. Oz, “Large N field theories,
string theory and gravity,” Phys. Rept. 323, 183 (2000) [hep-th/9905111].
[12] O. Aharony, O. Bergman, D. L. Jafferis and J. Maldacena, “N=6 superconformal Chern-
Simons-matter theories, M2-branes and their gravity duals,” JHEP 0810, 091 (2008)
[arXiv:0806.1218 [hep-th]].
[13] S. S. Gubser, I. R. Klebanov and A. M. Polyakov, “Gauge theory correlators from noncritical
string theory,” Phys. Lett. B 428, 105 (1998) [hep-th/9802109].
[14] E. Witten, “Anti-de Sitter space and holography,” Adv. Theor. Math. Phys. 2, 253 (1998)
[hep-th/9802150].
[15] I. Heemskerk and J. Polchinski, “Holographic and Wilsonian Renormalization Groups,”
JHEP 1106, 031 (2011) [arXiv:1010.1264 [hep-th]].
[16] T. Faulkner, H. Liu and M. Rangamani, “Integrating out geometry: Holographic Wilsonian
RG and the membrane paradigm,” JHEP 1108, 051 (2011) [arXiv:1010.4036 [hep-th]].
[17] E. Witten, “Quantum gravity in de Sitter space,” hep-th/0106109.
[18] A. Strominger, “The dS / CFT correspondence,” JHEP 0110, 034 (2001) [hep-th/0106113].
[19] S. Sachdev, “Quantum Phase Transitions”, 2nd Edition, Cambridge University Press (2011).
[20] E. P. Verlinde, “On the Origin of Gravity and the Laws of Newton,” JHEP 1104, 029 (2011)
[arXiv:1001.0785 [hep-th]].
[21] J. Zaanen, A.J. Beekman, “The emergence of gauge invariance: the stay-at-home gauge
versus local-global duality,” Annals of Physics 327(4):1146-1161 (2012) [arXiv:1108.2791
[cond-mat]].
[22] P. Breitenlohner and D. Z. Freedman, “Positive Energy in anti-De Sitter Backgrounds and
62
Gauged Extended Supergravity,” Phys. Lett. B 115, 197 (1982).
[23] K. S. Thorne, R. H. Price and D. A. Macdonald, “Black holes: the membrane paradigm,”
Yale University Press, New Haven 1986.
[24] E. Witten, “Anti-de Sitter space, thermal phase transition, and confinement in gauge theo-
ries,” Adv. Theor. Math. Phys. 2, 505 (1998) [arXiv:hep-th/9803131].
[25] G. W. Gibbons and S. W. Hawking, ”Action integrals and partition functions in quantum
gravity,” Physical Review D 15 (10): 2752
[26] J. W. York, “Role of conformal three-geometry in the dynamics of gravitation,” Physical
Review Letters 28 (16): 1082
[27] J. M. Maldacena, “TASI 2003 lectures on AdS/CFT,” hep-th/0309246.
[28] O. Aharony, J. Marsano, S. Minwalla, K. Papadodimas and M. Van Raamsdonk, “The Hage-
dorn - deconfinement phase transition in weakly coupled large N gauge theories,” Adv. Theor.
Math. Phys. 8, 603 (2004) [hep-th/0310285].
[29] S. W. Hawking and D. N. Page, “Thermodynamics Of Black Holes In Anti-De Sitter Space,”
Commun. Math. Phys. 87, 577 (1983).
[30] J. M. Maldacena, “Wilson loops in large N field theories,” Phys. Rev. Lett. 80, 4859 (1998)
[arXiv:hep-th/9803002].
[31] S. J. Rey and J. T. Yee, “Macroscopic strings as heavy quarks in large N gauge theory and
anti-de Sitter supergravity,” Eur. Phys. J. C 22, 379 (2001) [arXiv:hep-th/9803001].
[32] Andreas Brandhuber, Nissan Itzhaki, Jacob Sonnenschein, Shimon Yankielowicz, “Wilson
Loops, Confinement, and Phase Transitions in Large N Gauge Theories from Supergravity,”
JHEP 9806 (1998) 001 [arXiv:hep-th/9803263]
[33] J. Polchinski and M. J. Strassler, “Hard scattering and gauge/string duality,” Phys. Rev.
Lett. 88, 031601 (2002) [arXiv:hep-th/0109174].
[34] C. P. Herzog, “A holographic prediction of the deconfinement temperature,” Phys. Rev. Lett.
98, 091601 (2007) [arXiv:hep-th/0608151].
[35] G. T. Horowitz and R. C. Myers, “The AdS/CFT correspondence and a new positive energy
conjecture for general relativity,” Phys. Rev. D 59, 026005 (1999) [arXiv:hep-th/9808079].
[36] S. Surya, K. Schleich and D. M. Witt, “Phase transitions for flat AdS black holes,” Phys.
Rev. Lett. 86, 5231 (2001) [arXiv:hep-th/0101134].
[37] T. Nishioka, S. Ryu and T. Takayanagi, “Holographic Superconductor/Insulator Transition
63
at Zero Temperature,” JHEP 1003, 131 (2010) [arXiv:0911.0962 [hep-th]].
[38] G. T. Horowitz and B. Way, “Complete Phase Diagrams for a Holographic Superconduc-
tor/Insulator System,” JHEP 1011, 011 (2010) [arXiv:1007.3714 [hep-th]].
[39] P. Kovtun, D. T. Son and A. O. Starinets, “Viscosity in strongly interacting quantum field
theories from black hole physics,” Phys. Rev. Lett. 94, 111601 (2005) [hep-th/0405231].
[40] G. Policastro, D. T. Son and A. O. Starinets, “The Shear viscosity of strongly coupled N=4
supersymmetric Yang-Mills plasma,” Phys. Rev. Lett. 87, 081601 (2001) [hep-th/0104066].
[41] P. Kovtun, D. T. Son and A. O. Starinets, “Holography and hydrodynamics: Diffusion on
stretched horizons,” JHEP 0310, 064 (2003) [hep-th/0309213].
[42] A. Buchel and J. T. Liu, “Universality of the shear viscosity in supergravity,” Phys. Rev.
Lett. 93, 090602 (2004) [hep-th/0311175].
[43] J. Mas, “Shear viscosity from R-charged AdS black holes,” JHEP 0603, 016 (2006) [hep-
th/0601144].
[44] D. T. Son and A. O. Starinets, “Hydrodynamics of r-charged black holes,” JHEP 0603, 052
(2006) [hep-th/0601157].
[45] A. Buchel, J. T. Liu and A. O. Starinets, “Coupling constant dependence of the shear viscosity
in N=4 supersymmetric Yang-Mills theory,” Nucl. Phys. B 707, 56 (2005) [hep-th/0406264].
[46] J. Casalderrey-Solana, H. Liu, D. Mateos, K. Rajagopal and U. A. Wiedemann,
“Gauge/String Duality, Hot QCD and Heavy Ion Collisions,” arXiv:1101.0618 [hep-th].
[47] M. Brigante, H. Liu, R. C. Myers, S. Shenker and S. Yaida, “Viscosity Bound Violation in
Higher Derivative Gravity,” Phys. Rev. D 77, 126006 (2008) [arXiv:0712.0805 [hep-th]].
[48] D. T. Son and A. O. Starinets, “Minkowski space correlators in AdS / CFT correspondence:
Recipe and applications,” JHEP 0209, 042 (2002) [hep-th/0205051].
[49] D. T. Son and A. O. Starinets, “Viscosity, Black Holes, and Quantum Field Theory,” Ann.
Rev. Nucl. Part. Sci. 57, 95 (2007) [arXiv:0704.0240 [hep-th]].
[50] M. Parikh and F. Wilczek, “An Action for black hole membranes,” Phys. Rev. D 58, 064011
(1998) [gr-qc/9712077].
[51] M. Brigante, H. Liu, R. C. Myers, S. Shenker and S. Yaida, “The Viscosity Bound and
Causality Violation,” Phys. Rev. Lett. 100, 191601 (2008) [arXiv:0802.3318 [hep-th]].
[52] N. Iqbal and H. Liu, “Universality of the hydrodynamic limit in AdS/CFT and the membrane
paradigm,” Phys. Rev. D 79, 025023 (2009) [arXiv:0809.3808 [hep-th]].
64
[53] S. Caron-Huot, P. Kovtun, G. D. Moore, A. Starinets and L. G. Yaffe, JHEP 0612, 015
(2006) [hep-th/0607237].
[54] A. Nata Atmaja and K. Schalm, JHEP 1008, 124 (2010) [arXiv:0802.1460 [hep-th]].
[55] R. Baier, P. Romatschke, D. T. Son, A. O. Starinets and M. A. Stephanov, “Relativistic
viscous hydrodynamics, conformal invariance, and holography,” JHEP 0804, 100 (2008)
[arXiv:0712.2451 [hep-th]].
[56] S. Bhattacharyya, V. EHubeny, S. Minwalla and M. Rangamani, “Nonlinear Fluid Dynamics
from Gravity,” JHEP 0802, 045 (2008) [arXiv:0712.2456 [hep-th]].
[57] I. Fouxon and Y. Oz, “Conformal Field Theory as Microscopic Dynamics of Incompressible
Euler and Navier-Stokes Equations,” Phys. Rev. Lett. 101, 261602 (2008) [arXiv:0809.4512
[hep-th]].
[58] S. Bhattacharyya, S. Minwalla and S. R. Wadia, “The Incompressible Non-Relativistic
Navier-Stokes Equation from Gravity,” JHEP 0908, 059 (2009) [arXiv:0810.1545 [hep-th]].
[59] I. Bredberg, C. Keeler, V. Lysov and A. Strominger, “From Navier-Stokes To Einstein,”
arXiv:1101.2451 [hep-th].
[60] V. E. Hubeny, S. Minwalla and M. Rangamani, “The fluid/gravity correspondence,”
arXiv:1107.5780 [hep-th].
[61] J. D. Brown and J. W. York, “Quasilocal energy and conserved charges derived from the
gravitational action,” Phys. Rev. D 47:1407 (1993)
[62] B. W. A. Leurs, Z. Nazario, D. I. Santiago, J. Zaanen “Non-Abelian hydrodynamics and
the flow of spin in spin - orbit coupled substances,” Annals of Physics, Volume 323, Issue 4,
April 2008, arXiv:0705.2953[cond-mat]
[63] I. F. Herbut, V. Juricic and B. Roy, “Theory of interacting electrons on honeycomb lattice,”
Phys. Rev. B 79, 085116 (2009) arXiv: 0811.0610[cond-mat]
[64] C. P. Herzog, P. Kovtun, S. Sachdev and D. T. Son, “Quantum critical transport, duality,
and M-theory,” Phys. Rev. D 75, 085020 (2007) [hep-th/0701036].
[65] R. C. Myers, S. Sachdev and A. Singh, “Holographic Quantum Critical Transport without
Self-Duality,” Phys. Rev. D 83, 066017 (2011) [arXiv:1010.0443 [hep-th]].
[66] N. Iqbal, H. Liu and M. Mezei, “Semi-local quantum liquids,” JHEP 1204, 086 (2012)
[arXiv:1105.4621 [hep-th]].
[67] P. W. Anderson, “In Praise of Unstable Fixed Points: The Way Things Actually Work,”
65
arXiv:cond-mat/0201431
[68] S. S. Gubser and F. D. Rocha, “Peculiar properties of a charged dilatonic black hole in
AdS5,” Phys. Rev. D 81, 046001 (2010) [arXiv:0911.2898 [hep-th]].
[69] K. Goldstein, S. Kachru, S. Prakash and S. P. Trivedi, “Holography of Charged Dilaton Black
Holes,” JHEP 1008, 078 (2010) [arXiv:0911.3586 [hep-th]].
[70] C. Charmousis, B. Gouteraux, B. S. Kim, E. Kiritsis and R. Meyer, “Effective Holo-
graphic Theories for low-temperature condensed matter systems,” JHEP 1011, 151 (2010)
[arXiv:1005.4690 [hep-th]].
[71] G. T. Horowitz, “Introduction to Holographic Superconductors,” arXiv:1002.1722 [hep-th].
[72] G. T. Horowitz, J. E. Santos and D. Tong, “Optical Conductivity with Holographic Lattices,”
arXiv:1204.0519 [hep-th].
[73] D. van der Marel, et al., “power-law optical conductivity with a constant phase angle in
high Tc superconductors,” Nature 425 (2003) 271 [arXiv:cond-mat/0309172].
[74] Denis, Dalidovich, P. Phillips, “Nonlinear Transport Near a Quantum Phase Transition in
Two Dimensions,” Phys. Rev. Lett. vol. 93, 27004-1 (2004) [arXiv: cond-mat/0310129].
[75] S. A. Hartnoll and D. M. Hofman, “Locally critical umklapp scattering and holography,”
arXiv:1201.3917 [hep-th].
[76] S. S. Gubser, “Breaking an Abelian gauge symmetry near a black hole horizon,” Phys. Rev.
D 78, 065034 (2008) [arXiv:0801.2977 [hep-th]].
[77] S. A. Hartnoll, C. P. Herzog and G. T. Horowitz, “Building a Holographic Superconductor,”
Phys. Rev. Lett. 101, 031601 (2008) [arXiv:0803.3295 [hep-th]].
[78] S. A. Hartnoll, C. P. Herzog and G. T. Horowitz, “Holographic Superconductors,” JHEP
0812, 015 (2008) [arXiv:0810.1563 [hep-th]].
[79] J. Zaanen, “A modern, but way too short history of the theory of superconductivity at a
high temperature,” Chapter in the book “100 years of superconductivity” (eds. H. Rochalla
and P.H. Kes, Chapman and Hall) [ arXiv:1012.5461 [cond-mat.supr-con]]
[80] R. Ruffini and J. A. Wheeler “Introducing the Black Hole.” Physics Today 24, 30- 41, Jan.
1971.
[81] T. Faulkner, G. T. Horowitz, J. McGreevy, M. M. Roberts and D. Vegh, “Photoemission
’experiments’ on holographic superconductors,” JHEP 1003, 121 (2010) [arXiv:0911.3402
[hep-th]].
66
[82] S. S. Gubser and S. S. Pufu, “The Gravity dual of a p-wave superconductor,” JHEP 0811,
033 (2008) [arXiv:0805.2960 [hep-th]].
[83] M. M. Roberts and S. A. Hartnoll, “Pseudogap and time reversal breaking in a holographic
superconductor,” JHEP 0808, 035 (2008) [arXiv:0805.3898 [hep-th]].
[84] F. Benini, C. P. Herzog, R. Rahman and A. Yarom, “Gauge gravity duality for d-wave
superconductors: prospects and challenges,” JHEP 1011, 137 (2010) [arXiv:1007.1981 [hep-
th]].
[85] F. Denef and S. A. Hartnoll, “Landscape of superconducting membranes,” Phys. Rev. D 79,
126008 (2009) [arXiv:0901.1160 [hep-th]].
[86] J. P. Gauntlett, J. Sonner and T. Wiseman, “Holographic superconductivity in M-Theory,”
Phys. Rev. Lett. 103, 151601 (2009) [arXiv:0907.3796 [hep-th]].
[87] S. S. Gubser, C. P. Herzog, S. S. Pufu and T. Tesileanu, “Superconductors from Super-
strings,” Phys. Rev. Lett. 103, 141601 (2009) [arXiv:0907.3510 [hep-th]].
[88] M. Ammon, J. Erdmenger, M. Kaminski and P. Kerner, “Superconductivity from
gauge/gravity duality with flavor,” Phys. Lett. B 680, 516 (2009) [arXiv:0810.2316 [hep-
th]].
[89] D. Anninos, S. A. Hartnoll and N. Iqbal, “Holography and the Coleman-Mermin-Wagner
theorem,” Phys. Rev. D 82, 066008 (2010) [arXiv:1005.1973 [hep-th]].
[90] G. T. Horowitz, J. E. Santos and B. Way, “A Holographic Josephson Junction,” Phys. Rev.
Lett. 106, 221601 (2011) [arXiv:1101.3326 [hep-th]].
[91] E. Kiritsis and V. Niarchos, “Josephson Junctions and AdS/CFT Networks,” JHEP 1107,
112 (2011) [Erratum-ibid. 1110, 095 (2011)] [arXiv:1105.6100 [hep-th]].
[92] J. Bhattacharya, S. Bhattacharyya and S. Minwalla, “Dissipative Superfluid dynamics from
gravity,” JHEP 1104, 125 (2011) [arXiv:1101.3332 [hep-th]].
[93] C. P. Herzog, N. Lisker, P. Surowka and A. Yarom, “Transport in holographic superfluids,”
JHEP 1108, 052 (2011) [arXiv:1101.3330 [hep-th]].
[94] J. -H. She, B. J. Overbosch, Y. -W. Sun, Y. Liu, K. Schalm, J. A. Mydosh and J. Zaanen,
“Observing the origin of superconductivity in quantum critical metals,” Phys. Rev. B 84,
144527 (2011) [arXiv:1105.5377 [cond-mat.str-el]].
[95] E. Witten “Multitrace operators, boundary conditions, and AdS/CFT correspondence,” [hep-
th/0112258].
67
[96] J. -H. She and J. Zaanen, “BCS superconductivity in quantum critical metals,” Phys. Rev.
B 80, 184518 (2009).
[97] N. Iqbal and H. Liu, “Real-time response in AdS/CFT with application to spinors,” Fortsch.
Phys. 57, 367 (2009) [arXiv:0903.2596 [hep-th]].
[98] S. -S. Lee, “A Non-Fermi Liquid from a Charged Black Hole: A Critical Fermi Ball,” Phys.
Rev. D 79, 086006 (2009) [arXiv:0809.3402 [hep-th]].
[99] H. Liu, J. McGreevy and D. Vegh, “Non-Fermi liquids from holography,” Phys. Rev. D 83,
065029 (2011) [arXiv:0903.2477 [hep-th]].
[100] M. Cubrovic, J. Zaanen and K. Schalm, “String Theory, Quantum Phase Transitions and
the Emergent Fermi-Liquid,” Science 325, 439 (2009) [arXiv:0904.1993 [hep-th]].
[101] T. Faulkner, H. Liu, J. McGreevy and D. Vegh, “Emergent quantum criticality, Fermi sur-
faces, and AdS2,” Phys. Rev. D 83, 125002 (2011) [arXiv:0907.2694 [hep-th]].
[102] T. Faulkner, N. Iqbal, H. Liu, J. McGreevy and D. Vegh, “From Black Holes to Strange
Metals,” arXiv:1003.1728 [hep-th].
[103] J. Zaanen, F. Kruger, J.-H. She, D. Sadri, S. I. Mukhin “Pacifying the Fermi-liquid: battling
the devious fermion signs,” [arXiv:0802.2455 [cond-mat]].
[104] F. Kruger, J. Zaanen, “Fermionic quantum criticality and the fractal nodal surface” Phys.
Rev. B 78, 035104 (2008) [arXiv:0804.2161 [cond-mat.str-el]].
[105] J. Zaanen, B. J. Overbosch, “Mottness collapse and statistical quantum criticality” Phil.
Trans. R. Soc. A 369, 1599-1625 (2011) [arXiv:0911.4070 [cond-mat.str-el]]
[106] J. L. Petersen, Int. J. Mod. Phys. A 14, 3597 (1999) [hep-th/9902131].
[107] V. Balasubramanian, P. Kraus and A. E. Lawrence, Phys. Rev. D 59, 046003 (1999) [hep-
th/9805171].
[108] S.-S. Lee “Low energy effective theory of Fermi surface coupled with U(1) gauge field in 2+1
dimensions” Phys. Rev. B 80, 165102 (2009) [arXiv:0905.4532 [cond-mat.str-el]]
[109] M. A. Metlitski, S. Sachdev “Quantum phase transitions of metals in two spatial dimensions:
II. Spin density wave order” Phys.Rev.B82:075128,2010 [arXiv:1005.1288 [cond-mat.str-el]]
[110] E. Fradkin, S. A. Kivelson, M. J. Lawler, J. P. Eisenstein and A. P. Mackenzie, “Nematic
Fermi Fluids in Condensed Matter Physics”, Annual Review of Condensed Matter Physics
1, 153 (2010), [arXiv:0910.4166 [cond- mat.str-el]]
[111] S. Sachdev, private communication.
68
[112] J. Polchinski, In *Boulder 1992, Proceedings, Recent directions in particle theory* 235-274,
and Calif. Univ. Santa Barbara - NSF-ITP-92-132 (92,rec.Nov.) 39 p. (220633) Texas Univ.
Austin - UTTG-92-20 (92,rec.Nov.) 39 p [hep-th/9210046].
[113] R. Shankar, “Renormalization group approach to interacting fermions,” Rev. Mod. Phys. 66,
129 (1994).
[114] S. -X. Yang et al. “Proximity of the Superconducting Dome and the Quantum Criti-
cal Point in the Two-Dimensional Hubbard Model” Phys. Rev. Lett. 106, 047004 (2011)
[arXiv:1101.6050 [cond-mat.supr-con]] and references therein.
[115] V. Juricis, I. F. Herbut, G. W. Semenoff “Coulomb interaction at the metal-insulator critical
point in graphene” Phys.Rev.B80:081405,2009 [arXiv:0906.3513 [cond-mat.str-el]]
[116] J. A. Hertz, “Quantum critical phenomena,” Phys. Rev. B 14, 1165 (1976).
[117] A. V. Chubukov, D. Pines and J. Schmalian, in The physics of superconductors, Vol. 1 (eds
Bennemann, K. H. & Ketterson, J. B.) (Springer, 2004).
[118] E.-G. Moon and A. V. Chubukov, “Quantum-critical pairing with varying exponents,” J.
Low Temp. Phys. 161, 263 (2010) and references therein.
[119] J. T. Anderson and A. M. Goldman, “Experimental determination of the pair susceptibility
of a superconductor,” Phys. Rev. Lett. 25, 743 (1970).
[120] A. M. Goldman, “The order parameter susceptibility and collective modes of superconduc-
tors,” J. Supercond. Nov. Magn. 19, 317 (2006).
[121] R. A. Ferrell, “Fluctuations and the superconducting phase transition: II. Onset of Josephson
tunneling and paraconductivity of a junction,” J. Low Temp. Phys. 1, 423 (1969).
[122] D. J. Scalapino, “Pair tunneling as a probe of fluctuations in superconductors,” Phys. Rev.
Lett. 24, 1052 (1970).
[123] R. Contino and A. Pomarol, “Holography for fermions,” JHEP 0411, 058 (2004) [hep-
th/0406257].
[124] N. Bergeal, et al. “Pairing fluctuations in the pseudogap state of copper-oxide superconduc-
tors probed by the Josephson effect,” Nature Physics 4, 608 (2008).
[125] J. P. Gauntlett, J. Sonner and D. Waldram, “Universal fermionic spectral functions from
string theory,” Phys. Rev. Lett. 107, 241601 (2011) [arXiv:1106.4694 [hep-th]].
[126] J. P. Gauntlett, J. Sonner and D. Waldram, “Spectral function of the supersymmetry cur-
rent,” JHEP 1111, 153 (2011) [arXiv:1108.1205 [hep-th]].
69
[127] O. DeWolfe, S. S. Gubser and C. Rosen, “Fermi Surfaces in Maximal Gauged Supergravity,”
Phys. Rev. Lett. 108, 251601 (2012) [arXiv:1112.3036 [hep-th]].
[128] O. DeWolfe, S. S. Gubser and C. Rosen, “Fermi surfaces in N=4 Super-Yang-Mills theory,”
Phys. Rev. D 86, 106002 (2012) [arXiv:1207.3352 [hep-th]].
[129] S. S. Gubser and A. Nellore, “Ground states of holographic superconductors,” Phys. Rev. D
80, 105007 (2009) [arXiv:0908.1972 [hep-th]].
[130] G. T. Horowitz and M. M. Roberts, “Zero Temperature Limit of Holographic Superconduc-
tors,” JHEP 0911, 015 (2009) [arXiv:0908.3677 [hep-th]].
[131] N. Iqbal, H. Liu, M. Mezei and Q. Si, “Quantum phase transitions in holographic models
of magnetism and superconductors,” Phys. Rev. D 82, 045002 (2010) [arXiv:1003.0010 [hep-
th]].
[132] N. Iqbal, H. Liu and M. Mezei, “Quantum phase transitions in semi-local quantum liquids,”
arXiv:1108.0425 [hep-th].
[133] T. Faulkner and J. Polchinski, “Semi-Holographic Fermi Liquids,” arXiv:1001.5049 [hep-th].
[134] D. B. Kaplan, J. -W. Lee, D. T. Son and M. A. Stephanov, “Conformality Lost,” Phys. Rev.
D 80, 125005 (2009) [arXiv:0905.4752 [hep-th]].
[135] C. M. Varma, P. B. Littlewood, S. Schmitt-Rink, E. Abrahams and A. E. Ruckenstein,
“Phenomenology of the normal state of Cu-O high-temperature superconductors,” Phys.
Rev. Lett. 63, 1996 (1989).
[136] T. Hartman and S. A. Hartnoll, “Cooper pairing near charged black holes,” JHEP 1006,
005 (2010) [arXiv:1003.1918 [hep-th]].
[137] B. Pioline and J. Troost, “Schwinger pair production in AdS(2),” JHEP 0503, 043 (2005)
[hep-th/0501169].
[138] F. Massee et al. “Bilayer manganites: polarons in the midst of a metallic breakdown,” Nature
Physics 7, 978-982 (2011) [arXiv:1103.3804 [cond-mat.str-el]]
[139] M. Edalati, R. G. Leigh and P. W. Phillips, “Dynamically Generated Mott Gap from Holog-
raphy,” Phys. Rev. Lett. 106, 091602 (2011) [arXiv:1010.3238 [hep-th]].
[140] M. Edalati, R. G. Leigh, K. W. Lo and P. W. Phillips, “Dynamical Gap and Cuprate-like
Physics from Holography,” Phys. Rev. D 83, 046012 (2011) [arXiv:1012.3751 [hep-th]].
[141] P. Phillips “Mottness: Identifying the Propagating Charge Modes in doped Mott Insulators,”
Rev. Mod. Phys. vol. 82, 1719 (2010) [arXiv:1001.5270 [cond-mat.str-el]]
70
[142] J. Zaanen, A.J. Beekman “The emergence of gauge invariance: the stay-at-home gauge ver-
sus local-global duality” Annals of Physics 327(4):1146-1161 (2012) [arXiv:1108.2791 [cond-
mat.str-el]]
[143] S. S. Gubser and J. Ren, “Analytic fermionic Green’s functions from holography,” Phys. Rev.
D 86, 046004 (2012) [arXiv:1204.6315 [hep-th]].
[144] Y. Liu, K. Schalm, Y. -W. Sun and J. Zaanen, “Lattice potentials and fermions in holographic
non Fermi-liquids: hybridizing local quantum criticality,” arXiv:1205.5227 [hep-th].
[145] S. A. Hartnoll and A. Tavanfar, “Electron stars for holographic metallic criticality,” Phys.
Rev. D 83, 046003 (2011) [arXiv:1008.2828 [hep-th]].
[146] N. Iizuka, N. Kundu, P. Narayan and S. P. Trivedi, “Holographic Fermi and Non-Fermi
Liquids with Transitions in Dilaton Gravity,” arXiv:1105.1162 [hep-th].
[147] S. A. Hartnoll, D. M. Hofman and D. Vegh, “Stellar spectroscopy: Fermions and holographic
Lifshitz criticality, ” arXiv:1105.3197 [hep-th].
[148] S. A. Hartnoll and P. Petrov, “Electron star birth: A continuous phase transition at nonzero
density,” Phys. Rev. Lett. 106, 121601 (2011) [arXiv:1011.6469 [hep-th]].
[149] J. Erlich, E. Katz, D. T. Son and M. A. Stephanov, “QCD and a holographic model of
hadrons,” Phys. Rev. Lett. 95, 261602 (2005) [hep-ph/0501128].
[150] A. Karch, E. Katz, D. T. Son and M. A. Stephanov, “Linear confinement and AdS/QCD,”
Phys. Rev. D 74, 015005 (2006) [hep-ph/0602229].
[151] M. Cubrovic, Y. Liu, K. Schalm, Y. -W. Sun and J. Zaanen, “Spectral probes of the holo-
graphic Fermi groundstate: dialing between the electron star and AdS Dirac hair,” Phys.
Rev. D 84, 086002 (2011) [arXiv:1106.1798 [hep-th]].
[152] U. Gursoy and E. Kiritsis, “Exploring improved holographic theories for QCD: Part I,” JHEP
0802, 032 (2008) [arXiv:0707.1324 [hep-th]].
[153] U. Gursoy, E. Kiritsis and F. Nitti, “Exploring improved holographic theories for QCD: Part
II,” JHEP 0802, 019 (2008) [arXiv:0707.1349 [hep-th]].
[154] J. Erdmenger, N. Evans, I. Kirsch and E. Threlfall, “Mesons in Gauge/Gravity Duals - A
Review,” Eur. Phys. J. A 35, 81 (2008) [arXiv:0711.4467 [hep-th]].
[155] J. Casalderrey-Solana, H. Liu, D. Mateos, K. Rajagopal and U. A. Wiedemann,
“Gauge/String Duality, Hot QCD and Heavy Ion Collisions,” arXiv:1101.0618 [hep-th].
[156] T. Nishioka, S. Ryu and T. Takayanagi, “Holographic Superconductor/Insulator Transition
71
at Zero Temperature,” JHEP 1003, 131 (2010) [arXiv:0911.0962 [hep-th]].
[157] G. T. Horowitz and B. Way, “Complete Phase Diagrams for a Holographic Superconduc-
tor/Insulator System,” JHEP 1011, 011 (2010) [arXiv:1007.3714 [hep-th]].
[158] M. Cubrovic, J. Zaanen and K. Schalm, “Constructing the AdS Dual of a Fermi Liquid: AdS
Black Holes with Dirac Hair,” JHEP 1110, 017 (2011) [arXiv:1012.5681 [hep-th]].
[159] S. Sachdev, “A model of a Fermi liquid using gauge-gravity duality,” Phys. Rev. D 84, 066009
(2011) [arXiv:1107.5321 [hep-th]].
[160] A. Allais, J. McGreevy and S. J. Suh, “A quantum electron star,” arXiv:1202.5308 [hep-th].
[161] S. A. Hartnoll and P. Petrov, “Electron star birth: A continuous phase transition at nonzero
density,” Phys. Rev. Lett. 106, 121601 (2011) [arXiv:1011.6469 [hep-th]].
[162] S. Nakamura, H. Ooguri and C. -S. Park, “Gravity Dual of Spatially Modulated Phase,”
Phys. Rev. D 81, 044018 (2010) [arXiv:0911.0679 [hep-th]].
[163] T. Sakai and S. Sugimoto, “Low energy hadron physics in holographic QCD,” Prog. Theor.
Phys. 113, 843 (2005) [hep-th/0412141].
[164] A. Karch and E. Katz, “Adding flavor to AdS / CFT,” JHEP 0206, 043 (2002) [hep-
th/0205236].
[165] S. -J. Rey, “String theory on thin semiconductors: Holographic realization of Fermi points
and surfaces,” Prog. Theor. Phys. Suppl. 177, 128 (2009) [arXiv:0911.5295 [hep-th]].
[166] A. Karch and L. Randall, “Localized gravity in string theory,” Phys. Rev. Lett. 87, 061601
(2001) [hep-th/0105108].
[167] A. Karch and A. O’Bannon, “Metallic AdS/CFT,” JHEP 0709, 024 (2007) [arXiv:0705.3870
[hep-th]].
[168] S. A. Hartnoll, J. Polchinski, E. Silverstein and D. Tong, “Towards strange metallic holog-
raphy,” JHEP 1004, 120 (2010) [arXiv:0912.1061 [hep-th]].
[169] A. Karch, D. T. Son and A. O. Starinets, “Zero Sound from Holography,” arXiv:0806.3796
[hep-th].
[170] M. Kulaxizi and A. Parnachev, “Comments on Fermi Liquid from Holography,” Phys. Rev.
D 78, 086004 (2008) [arXiv:0808.3953 [hep-th]].
[171] R. A. Davison and A. O. Starinets, “Holographic zero sound at finite temperature,” Phys.
Rev. D 85, 026004 (2012) [arXiv:1109.6343 [hep-th]].
[172] O. Bergman, N. Jokela, G. Lifschytz and M. Lippert, “Quantum Hall Effect in a Holographic
72
Model,” JHEP 1010, 063 (2010) [arXiv:1003.4965 [hep-th]].
[173] S. Kachru, A. Karch and S. Yaida, “Holographic Lattices, Dimers, and Glasses,” Phys. Rev.
D 81, 026007 (2010) [arXiv:0909.2639 [hep-th]].
[174] S. Kachru, A. Karch and S. Yaida, “Adventures in Holographic Dimer Models,” New J. Phys.
13, 035004 (2011) [arXiv:1009.3268 [hep-th]].
[175] H. Ooguri and C. -S. Park, “Holographic End-Point of Spatially Modulated Phase Transi-
tion,” Phys. Rev. D 82, 126001 (2010) [arXiv:1007.3737 [hep-th]].
[176] A. Donos and J. P. Gauntlett, “Holographic striped phases,” JHEP 1108, 140 (2011)
[arXiv:1106.2004 [hep-th]].
[177] O. Bergman, N. Jokela, G. Lifschytz and M. Lippert, “Striped instability of a holographic
Fermi-like liquid,” JHEP 1110, 034 (2011) [arXiv:1106.3883 [hep-th]].
[178] A. Donos and J. P. Gauntlett, “Holographic helical superconductors,” JHEP 1112, 091
(2011) [arXiv:1109.3866 [hep-th]].
[179] M. M. Wolf, “Violation of the entropic area law for Fermions,” Phys. Rev. Lett. 96, 010404
(2006) [quant-ph/0503219].
[180] S. Ryu and T. Takayanagi, “Holographic derivation of entanglement entropy from
AdS/CFT,” Phys. Rev. Lett. 96, 181602 (2006) [hep-th/0603001].
[181] N. Ogawa, T. Takayanagi and T. Ugajin, “Holographic Fermi Surfaces and Entanglement
Entropy,” JHEP 1201, 125 (2012) [arXiv:1111.1023 [hep-th]].
[182] L. Huijse, S. Sachdev and B. Swingle, “Hidden Fermi surfaces in compressible states of
gauge-gravity duality,” Phys. Rev. B 85, 035121 (2012) [arXiv:1112.0573 [cond-mat.str-el]].
[183] E. Shaghoulian, “Holographic Entanglement Entropy and Fermi Surfaces,” JHEP 1205, 065
(2012) [arXiv:1112.2702 [hep-th]].
[184] X. Dong, S. Harrison, S. Kachru, G. Torroba and H. Wang, “Aspects of holography for
theories with hyperscaling violation,” arXiv:1201.1905 [hep-th].
[185] N. Iqbal and H. Liu, “Luttinger’s Theorem, Superfluid Vortices, and Holography,”
arXiv:1112.3671 [hep-th].
73